# contemplating_realworld_object_classification__dd708fd3.pdf Published as a conference paper at ICLR 2021 CONTEMPLATING REAL-WORLD OBJECT CLASSIFICATION Ali Borji HCL America aliborji@gmail.com Deep object recognition models have been very successful over benchmark datasets such as Image Net. How accurate and robust are they to distribution shifts arising from natural and synthetic variations in datasets? Prior research on this problem has primarily focused on Image Net variations (e.g., Image Net V2, Image Net-A). To avoid potential inherited biases in these studies, we take a different approach. Specifically, we reanalyze the Object Net dataset1 recently proposed by Barbu et al. containing objects in daily life situations. They showed a dramatic performance drop of the state of the art object recognition models on this dataset. Due to the importance and implications of their results regarding the generalization ability of deep models, we take a second look at their analysis. We find that applying deep models to the isolated objects, rather than the entire scene as is done in the original paper, results in around 20-30% performance improvement. Relative to the numbers reported in Barbu et al., around 10-15% of the performance loss is recovered, without any test time data augmentation. Despite this gain, however, we conclude that deep models still suffer drastically on the Object Net dataset. We also investigate the robustness of models against synthetic image perturbations such as geometric transformations (e.g., scale, rotation, translation), natural image distortions (e.g., impulse noise, blur) as well as adversarial attacks (e.g., FGSM and PGD-5). Our results indicate that limiting the object area as much as possible (i.e., from the entire image to the bounding box to the segmentation mask) leads to consistent improvement in accuracy and robustness. Finally, through a qualitative analysis of Object Net data, we find that i) a large number of images in this dataset are hard to recognize even for humans, and ii) easy (hard) samples for models match with easy (hard) samples for humans. Overall, our analyses show that Objec Net is still a challenging test platform for evaluating the generalization ability of models. Code and data are available at https://github.com/aliborji/Object Net Reanalysis.git2. 1 INTRODUCTION Object recognition3 can be said to be the most basic problem in vision sciences. It is required in the early stages of visual processing before a system, be it a human or a machine, can accomplish other tasks such as searching, navigating, or grasping. Application of a convolutional neural network architecture (CNN) known as Le Net (Le Cun et al., 1998), albeit with new bells and whistles (Krizhevsky et al., 2012), revolutionized not only computer vision but also several other areas. With the initial excitement gradually damping, researchers have started to study the shortcomings of deep models and question their generalization ability. From prior research, we already know that CNNs: a) lack generalization to out of distribution samples (e.g., Recht et al. (2019); Barbu et al. (2019); Shankar et al. (2020); Taori et al. (2020); Koh et al. (2020)). Even after being exposed to many different instances of the same object category, they fail to fully capture the concept. In stark contrast, humans can generalize from only few examples (a.k.a few-shot learning), b) perform poorly when applied to transformed versions of the same object. In other words, they 1https://objectnet.dev/ 2See https://openreview.net/forum?id=Q4EUyw JIkqr for reviews and discussions. A prelimnary version of this work has been published in Arxiv (Borji, 2020). 3Classification of an object appearing lonely in an image. For images containing multiple objects, object localization or detection is required first. Published as a conference paper at ICLR 2021 are not invariant to spatial transformations (e.g., translation, in-plane and in-depth rotation, scale) as shown in (Azulay & Weiss, 2019; Engstrom et al., 2019; Fawzi & Frossard, 2015), as well as noise corruptions (Hendrycks & Dietterich, 2019; Geirhos et al., 2018b), and c) are vulnerable to imperceptible adversarial image perturbations (Szegedy et al., 2013; Goodfellow et al., 2014; Nguyen et al., 2015). Majority of these works, however, have used either the Image Net dataset or its variations, and thus might be biased towards Image Net characteristics. Utilizing a very challenging dataset that has been proposed recently, known as Object Net (Barbu et al., 2019), here we seek to answer how well the state of the art CNNs generalize to real world object recognition scenarios. We also explore the role of spatial context in object recognition and answer whether it is better to use cropped objects (using bounding boxes) or segmented objects to achieve higher accuracy and robustness. Furthermore, we study the relationship between object recognition, scene understanding, and object detection. These are important problems that have been less explored. Several datasets have been proposed for training and testing object recognition models, and to study their generalization ability (e.g., Image Net by Deng et al. (2009), Places by Zhou et al. (2017), CIFAR by Krizhevsky et al. (2009), NORB by Le Cun et al. (2004), and i Lab20M by Borji et al. (2016)). As the most notable one, Image Net dataset has been very instrumental for gauging the progress in object recognition over the past decade. A large number of studies have tested new ideas by training deep models on Image Net (from scratch), or by finetuning pre-trained (on Image Net) classification models on other datasets. With the Image Net being retired, the state of the object recognition problem remains unclear. Several questions such as out of distribution generalization, superhuman performance (He et al., 2016) and invariance to transformations persist. To rekindle the discourse, recently Barbu et al. (2019) introduced the Object Net dataset which according to their claim has less bias than other recognition datasets4. This dataset is supposed to be used solely as a test set and comes with a licence that disallows the researchers to finetune models on it. Images are pictured by Mechanical Turk workers using a mobile app in a variety of backgrounds, rotations, and imaging viewpoints. Object Net contains 50,000 images across 313 categories, out of which 113 are in common with Image Net categories. Astonishingly, Barbu et al. found that the state of the art object recognition models perform drastically lower on Object Net compared to their performance on Image Net (about 40-45% drop). Our principal goal here it to revisit the Barbu et al. s analysis and measure the actual performance drop on Object Net compared to Image Net. To this end, we limit our analysis to the 113 overlapped categories between the two datasets. We first annotate the objects in the Object Net scenes by drawing boxes around them. We then apply a number of deep models on these object boxes and find that models perform significantly better now, compared to their performance on the entire scene (as is done in Barbu et. al). Interestingly, and perhaps against the common belief, we also find that training and testing models on segmented objects, rather than the object bounding box or the full image, leads to consistent improvement in accuracy and robustness over a range of classification tasks and image transformations (geometric, natural distortions, and adversarial attacks). Lastly, we provide a qualitative (and somewhat anecdotal) analysis of extreme cases in object recognition for humans and machines. 2 RELATED WORK Robustness against synthetic distribution shifts. Most research on assessing model robustness has been focused on synthetic image perturbations (e.g., spatial transformations, noise corruptions, simulated weather artifacts, temporal changes (Gu et al., 2019), and adversarial examples) perhaps because it is easy to precisely define, implement, and apply them to arbitrary images. While models have improved significantly in robustness to these distribution shifts (e.g., Zhang (2019); Zhang et al. (2019); Cohen & Welling (2016)), they are still not as robust as humans. Geirhos et al. (2018b) showed that humans are more tolerant against image manipulations like contrast reduction, additive noise, or novel eidolon-distortions than models. Further, humans and models behave differently (witnessed by different error patterns) as the signal gets weaker. Zhu et al. (2016) contrast the influence of the foreground object and image background on the performance of humans and models. Robustness against natural distribution shifts. Robustness on real data is a clear challenge for deep neural networks. Unlike synthetic distribution shifts, it is difficult to define distribution shifts that occur naturally in the real-world (such as subtle changes in scene composition, object types, and lighting conditions). Recht et al. (2019) closely followed the original Image Net creation process 4Object Net dataset, however, has it own biases. It consists of indoor objects that are available to many people, are mobile, are not too large, too small, fragile or dangerous. Published as a conference paper at ICLR 2021 Figure 1: Sample images from the Object Net dataset along with our manually annotated object bounding boxes from Chairs, Teapots and T-shirts categories. The leftmost column shows three chair examples from the Image Net dataset. Image Net scenes often have a single isolated object in them whereas images in the Object Net dataset contain multiple objects. Further, Object Net objects cover a wider range of variation in contrast, rotation, scale, and occlusion compared to Image Net objects (See arguments in Barbu et al. (2019)). In total, we annotated 18,574 images across 113 categories in common between the two datasets. This figure is modified from Figure 2 in Barbu et al. (2019). to build a new test set called Image Net V2. They reported a performance gap of about 11% (top-1 acc.) between the performance of the best deep models on this dataset and the original test set. Similar observations have been made by Shankar et al. (2020). By evaluating 204 Image Net models in 213 different test conditions, Taori et al. (2020) found that a) current synthetic robustness does not imply natural robustness. In other words, robustness measures for synthetic distribution shifts are weakly predictive of robustness on the natural distribution shifts, b) robustness measurements should control for accuracy since higher robustness can sometimes be explained by the higher accuracy on a standard unperturbed test set, and c) training models on larger and more diverse data improves robustness but does not lead to full closure of the performance gap. A comprehensive benchmark of distribution shifts in the wild, known as WILDS, has recently been published by Koh et al. (2020), encompassing different data modalities including vision. In D Amour et al. (2020), authors regard underspecification a major challenge to the credibility and generalization of modern machine learning pipelines. An ML pipeline is underspecified when it returns models that perform very well on held-out test sets during training but perform poorly at deployment time. Contextual interference. Context plays a significant role in pattern recognition and visual reasoning (e.g., Bar (2004); Torralba & Sinha (2001); Rabinovich et al. (2007); Heitz & Koller (2008); Galleguillos & Belongie (2010)). The extent to which visual context is being used by deep models is still unclear. Unlike models, humans are very good at exploiting context when it is helpful and discard it when it causes ambiguity. In other words, deep models do not understand what is the foreground object and what constitutes the background5. Nagarajan et al. (2020) mention that ML models utilize features (e.g., image background) which are spuriously correlated with the label during training. This makes them fragile at the test time when statistics slightly differ. As we argue here, this is one of the main reasons why deep models are so vulnerable to geometric and adversarial perturbations. Geirhos et al. (2020) have studied this phenomenon under the shortcut learning terminology from a broader perspective. Insights from human vision. CNNs turn out to be good models of human vision and can explain the first feed-forward sweep of information (See Kriegeskorte (2015) for a review). They, however, differ from human visual processing in several important ways. Current object recognition methods do not rely on segmentation, whereas figure-ground segmentation plays a significant role in human vision, in particular for the encoding of spatial relations between 3D object parts (Biederman, 1987; Serre, 2019). Some computer vision works, predating deep learning, have also shown that pre-segmenting the image before applying the recognition algorithms, improves the accuracy (Malisiewicz & Efros, 2007; Rabinovich et al., 2007; Rosenfeld & Weinshall, 2011). Unlike the human vision system, CNNs are hindered drastically in crowded scenes (e.g., Volokitin et al. (2017)). CNNs rely more on texture whereas humans pay more attention to shape (Geirhos et al., 2018a). Utilizing minimal recognizable images, Ullman et al. (2016) argued that the human visual system uses features and processes that are not used by current deep models. 5As an example, consider a model that is trained to classify camels vs. cows, with camels always shown in sandy backgrounds and cows shown against grassy backgrounds. Although such a model does well during training, it gets confused when presented with cows in sandy backgrounds at test time (Beery et al., 2018). See also Rosenfeld et al. (2018) for another example in the context of object detection Published as a conference paper at ICLR 2021 Figure 2: Performance of deep object recognition models on Object Net dataset. Results of our analysis by applying Alex Net, VGG-19, and Res Net-152 models to object bounding boxes (i.e., isolated objects) are shown in blue (overlaid in Fig. 1 from the Object Net paper). Feeding object boxes to models instead of the entire scene improves the accuracy about 10-15%. The right panel shows the results using our code (thus leading to a fair comparison). Performance improvement is more significant now which is 20-30%. We did not use test time data augmentation here. Our analysis Object Net paper < Recognizers by year > Object Net Top-5 (box) Object Net Top-1 (box) 25-35% performance drop 40-45% performance drop Using our code 3 EXPERIMENTS AND RESULTS 3.1 ACCURACY AND ROBUSTNESS AGAINST NATURAL DISTRIBUTION SHIFTS A critic of Barbu et al. (2019). Barbu et al. s work is a great contribution to the field to answer how well object recognition models generalize to the real-world circumstances and to control for biases in data collection. It, however, suffers from a major shortcoming that is making no distinction between object detection and object recognition . This confusion brings along several concerns: 1. They use the term object detector to refer to object recognition models. Object detection and object recognition are two distinct, yet related, tasks. Each one has its own models, datasets, evaluation measures, and inductive biases. For example, as shown in Fig. 1, images in object recognition datasets (e.g., Image Net) often contain a single object, usually from a closeup view, whereas scenes in object detection datasets (e.g., MS COCO (Lin et al., 2014), Open Images (Kuznetsova et al., 2018)) usually have multiple objects. Objects in the detection datasets vary more in some parameters such as occlusion and size. For instance, there is a larger variation in object scale in detection datasets (Singh & Davis, 2018). This discussion also relates to the distinction between scene understanding and object recognition . To understand a complex scene, as humans we look around, fixate on individual objects to recognize them, and accumulate information over fixations to perform more complex tasks such as answering a question or describing an event. To avoid biases in recognition datasets (e.g., typical scales or object views), we propose to (additionally) use detection datasets to study object recognition. We will discuss this further in Section 4. 2. Instead of applying models to isolated objects, Barbu et al. apply them to cluttered scenes containing multiple objects. Unlike Image Net where the majority of images include only a single object, Object Net images have multiple objects in them and are often more cluttered. Therefore, the drop in performance of models on Object Net can be merely due to the fact that pretrained models on Image Net have been trained on individual objects. 3. In addition to top-1 accuracy, Barbu et al. also report top-5 accuracy. One might argue that this may suffice in dealing with scenes containing multiple objects. Top-5 accuracy was first introduced in Russakovsky et al. (2015) to remedy the issues with the top-1 accuracy. The latter can be overly stringent by penalizing predictions that appear in the image but do not correspond to the target label. Top-5 accuracy itself, however, has two shortcomings. First, a model can still be penalized if all of the five guesses exist in the image, but none is the image label. Both scores fall short in addressing the images with counter-intuitive labels (e.g., when non-salient objects are labeled; Appx. E). Second, on fine-grained classification tasks (Image Net has several fine-grained classes e.g., dogs), allowing five Published as a conference paper at ICLR 2021 predictions can make certain class distinctions trivial (Shankar et al., 2020). For example, there are five turtles in the Image Net class hierarchy (mud turtle, box turtle, loggerhead turtle, leatherback turtle, and terrapin) that are difficult to distinguish. A classifier may trick the score by generating all of these labels for a turtle image to ensure it predicts the correct label. Shankar et al. proposed to use multi-label accuracy as an alternative to top-5 score. Each image has a set of target labels (i.e., multi-label annotations). A prediction is marked correct if it corresponds to any of the target labels for that image. This score, however, may favor a model that generates correct labels but may confuse the locations over a model that is spatially more precise but misses some objects (See also Beyer et al. (2020)). Regardless, since multi-label annotations for Object Net are not available, we report both top-1 and top-5 scores when feeding isolated objects to models. Bounding box annotation. The 113 object categories in the Object Net dataset, overlapped with the Image Net, contain 18,574 images in total. On this subset, the average number of images per category is 164.4 (min=55, max=284). Fig. 8 in Appx. A shows the distribution of the number of images per category on this dataset (envelope and dish drying rack are the most and least frequent objects, respectively). We drew a bounding box around the object corresponding to the category label of each image. If there were multiple nearby objects from the same category (e.g., chairs around a table), we tried to include all of them in the bounding box. Some example scenes and their corresponding bounding boxes are given in Fig. 1. Appx. H shows more stats on Object Net. Object recognition results. We employ six widely-used state of the art deep neural networks including Alex Net (Krizhevsky et al., 2012), VGG-19 (Simonyan & Zisserman, 2014), Goog Le Net (Szegedy et al., 2015), Res Net-152 (He et al., 2016), Inception-v3 (Szegedy et al., 2016)6, and MNASNet (Tan et al., 2019). Alex Net, VGG-19, and Res Net-152 have also been used in the Object Net paper (Barbu et al., 2019). We use the Py Torch implementation of these models7. Since the code from the Object Net paper is unavailable (at the time of preparing this work), in addition to applying models to bounding boxes and plotting the results on top of the results from the Object Net paper, we also run our code on both the bounding boxes and the full images. This allows a fair comparison and helps mitigate possible inconsistency in data processing methods (e.g., different data normalization schemes or test time data augmentation such as rotation, scale, color jittering, cropping, etc.). Fig. 2 shows an overlay of our results in Fig. 1 from the Object Net paper. As can be seen, applying models to the object bounding box instead of the entire scene improves the accuracy about 10-15%. Although the gap is narrower now, models still significantly underperform on Object Net than the Image Net dataset. Using our code, the improvement going from full image to bounding boxes is around 20-30% across all tested models (the right panel in Fig. 2). Our results using the full image are lower than Barbu et al. s results using the full image (possibly because we do not utilize data augmentation). This relative difference entails that applying their code to bounding boxes will likely improve the performance beyond 10% that we obtained here. Assuming 25% gain in performance on top of their best results when using boxes, will still not close the performance gap which indicates that Object Net remains a challenging dataset for testing object recognition models. Breakdown of accuracy over the 113 categories is shown in Appx. B (Figs. 9 & 10 over isolated objects and Figs. 11 & 12 over the full image). Interestingly, in both cases, almost all models, except Goog Le Net on isolated objects and Alex Net on the full image, perform the best over the safety pin category. Inspecting the images from this class, we found that they have a single safety pin often held by a person (perhaps about the same distance from the camera thus similar scales). The same story is true about the banana class which is the second easiest category using the bounding boxes. This object becomes much harder to recognize when using the full image (26.88% vs. 70.3% using boxes) which highlights the benefit of applying models to isolated objects rather than scenes. 3.2 ACCURACY AND ROBUSTNESS AGAINST SYNTHETIC DISTRIBUTION SHIFTS 3.2.1 ROBUSTNESS AGAINST COMMON IMAGE CORRUPTIONS Previous work has shown that Image Net-trained CNNs generalize poorly over a wide range of image distortions (e.g., Hendrycks & Dietterich (2019); Azulay & Weiss (2019); Dodge & Karam (2017)). These works, however, have applied CNNs to the whole scene. Here, we 6Barbu et al. have used Inception-v4. 7https://pytorch.org/docs/stable/torchvision/models.html Published as a conference paper at ICLR 2021 Figure 3: Left two panels) Average top-1 and top-5 accuracy of models over 1130 images from the Object Net dataset corrupted by 14 natural image distortions. Right two panels) Average classification accuracy per each of the 113 categories (left) and each of the 14 distortion types (right). In all cases, applying models to the isolated object (using bounding boxes) leads to higher robustness than applying them to the full image. ask whether applying the models to the bounding boxes can improve robustness against image distortions. Following Hendrycks & Dietterich (2019), we systematically test how model accuracy degrades if images are corrupted by 14 different types of distortions including Gaussian noise, shot noise, impulse noise, defocus blur, glass blur, motion blur, zoom blur, snow, frost, fog, brightness, contrast, elastic transform, and JPEG compression at 3 levels of corruption severity. Fig. 36 (Appx. F) shows sample images along with their distortions. Ten images from each of the 113 categories of Object Net (1130 images in total) were fed to three models including VGG-19, Inception-v3, and Res Net-152. Aggregate results over the full image and the object bounding box (both resized to 224 224 pixels) are shown in Fig. 3. All three models are more robust when applied to the object bounding box than the full image at all corruption levels, using both top-1 and top-5 scores (left two panels). Among models, Res Net-152 performs better and is the most robust model. It is followed by the Inception-v3 model. For nearly all of the 113 object categories, using bounding boxes leads to higher robustness than using the full image (the third panel). Similarly, using bounding boxes results in higher robustness against all distortion types (the right-most panel). Across distortion types, shown in Figs. 37 & 38 (Appx. F), Res Net-152 consistently outperforms the other two models at all severity levels, followed by Inception-v3. It seems that models are hindered more by impulse noise, frost, zoom blur, and snow distortions. The top-1 accuracy at severity level 2 on these distortions is below 20%. Overall, we conclude that limiting the object area only to the bounding box leads not only to higher prediction accuracy but also to higher robustness against image distortions. Extrapolating this approach, can we improve robustness by shrinking the object region even further by using the segmentation masks? We will thoroughly investigate this question in the next subsections. 3.2.2 ROBUSTNESS AGAINST ADVERSARIAL PERTURBATIONS Despite being very accurate, CNNs are highly vulnerable to adversarial inputs (Szegedy et al., 2013; Goodfellow et al., 2014). These inputs are crafted carefully and maliciously by adding small imperceptible perturbations to them (e.g., altering the value of a pixel up to 8 units under the ℓ -norm; pixels in the range [0, 255]). Here we apply the Image Net pretrained models to 1130 images that were selected above. The models are tested against the Fast Gradient Sign Method (FGSM) (Goodfellow et al., 2014) at two perturbation budgets in the untargeted white-box setting. Full Image Bounding Box Model ϵ = 2/255 ϵ = 8/255 ϵ = 2/255 ϵ = 8/255 VGG-19 0.53/2.48 0.09/0.71 3.27/10.44 1.06/5.66 Inception-v3 3.18/10.62 1.42/4.42 9.03/25.13 4.87/15.6 Res Net-152 2.39/10.34 1.15/4.96 10.62/26.64 6.64/19.73 Table 1: Robust accuracy (top-1/top-5) against FGSM attack. Table. 1 shows the results. We find that models are more resilient against the FGSM attack when applied to the bounding box than the full image. While the input size is the same in both cases (224 224), the adversary has more opportunity to mislead the classifier in the full image case since a larger fraction of pixels play an insignificant role in the decisions made by the network. This aligns with observations from the visualization tools (e.g., Selvaraju et al. (2017)) revealing that CNNs indeed rely only on a small subset of image pixels to elicit a decision. One might argue that the lower robustness on the full images could be due to training and test discrepancy (i.e., training models on single objects and applying them to the entire scene). To address this, in the next subsection we train and test models in the same condition. Published as a conference paper at ICLR 2021 3.3 THE INFLUENCE OF THE SURROUNDING CONTEXT ON ROBUSTNESS Fashion MNIST Figure 4: The effect of background subtraction (a.k.a foreground detection) on adversarial robustness (here against the FGSM attack). Two models are trained and tested on clean and noisy data from MNIST (left) and Fashion MNIST (right) datasets. In the noise case, the object is overlaid in a white noise field (no noise on the object itself). Despite a large body of literature on whether and how much visual context benefits CNNs in terms of accuracy and robustness8, the matter has not been settled yet (e.g., Bar (2004); Torralba & Sinha (2001); Rabinovich et al. (2007); Rosenfeld et al. (2018); Heitz & Koller (2008); Divvala et al. (2009); Zhu et al. (2016); Xiao et al. (2020); Malisiewicz & Efros (2007)). To study how context surrounding an object impacts model accuracy and robustness in more detail, we conducted two experiments. In the first one, we trained two CNNs (2 conv layers, each followed by a pooling layer and 2 final fc layers) on MNIST and Fashion MNIST datasets, for which it is easy to derive the foreground masks (Figs. 39 & 40; Appx. G). CNNs were trained on either the original clean images or the foreground objects placed on a white noise background. We then tested the models against the FGSM attack w/o background subtraction. With background subtraction, we essentially assume that the adversary has access only to the foreground object (i.e., effectively removing the perturbations that fall on the background). As results in Fig. 4 show, background subtraction improves the robustness substantially using both models and over both datasets. To examine whether the above conclusion generalizes to more complex natural scenes, we ran a second experiment. First, we selected images from ten classes of the MS COCO dataset including chair, car, book, bottle, dinning table, umbrella, boat, motorcycle, sheep, and cow. Objects from these classes come with a segmentation mask (one object per image; 100 images per category; 1000 images in total). Around 32.7% of the image pixels fall inside the object bounding box and around 58.1% of the bounding box pixels fall inside the object mask. Fig. 5 shows a sample chair alongside its bounding box and its segmentation mask. Bounding Box Seg. Mask Figure 5: Model accuracy against adversarial perturbations (left) and noise corruptions (middle). The right panel shows a sample chair image along with its bounding box and segmentation mask. We then trained three Res Net-18 models (finetuned on Image Net), one per each input type: 1) full image, 2) bounding box, and 3) segmented object (placed in a dark background). Models were trained on 70 images per category (700 in total) for 10 epochs and were then tested on the remaining 30 images per category. An attempt was made to tune the parameters to attain the best test accuracy in each case (e.g., by avoiding overfitting). The test accuracy9 of models in order are 66.9%, 78%, and 80.3%. One reason behind lower prediction accuracy using boxes might be because multiple objects may fit inside 8Majority of such works are focused on model accuracy 9 Taori et al. (2020) argue that robustness scores should control for accuracy as more predictive models in general are more robust. To avoid this issue we used models that have about the same standard accuracy. Published as a conference paper at ICLR 2021 Figure 6: Performance (top-1) of the Res Net-18 model against geometric transformations. the bounding box (e.g., for elongated objects such as broom). Model performance against FGSM and ℓ PGD-5 (Projected Gradient Descent by Madry et al. (2017)) adversarial attacks are shown in Fig. 5 (left panel). We observe that training models on segmented objects leads to higher adversarial robustness against both types of attacks. The improvement is more pronounced at higher perturbations. We also considered a condition in which we masked the perturbations that fall on the background, denoted as Seg. Mask + FG in the figure. We noticed even higher robustness against the attacks by removing the background perturbations. These results encourage using foreground detection as an effective adversarial defense. The middle panel in Fig. 5 shows model robustness against noise corruptions (averaged over the 14 distortions used in Section 3.2.1). Here again, we find that using segmentation masks leads to higher robustness compared to the full image and object boxes. Seg. Mask + FG leads to the best robustness among the input types. While it might be hard to draw a general conclusion regarding the superiority of the segmentation masks over bounding boxes in object recognition accuracy, our investigation suggests that using masks leads to a significant boost in adversarial robustness with little or no drop in standard accuracy. Our results offer an upper bound in the utility of segmentation masks in robustness. More work is needed to incorporate this feat in CNNs (i.e., using attention). 3.3.1 ROBUSTNESS AGAINST GEOMETRIC TRANSFORMATIONS We also tested the Res Net-18 model (i.e., trained over the full image, the bounding box, and the segmented object on Object Net; as above) against three geometric transformations including scaling, in-plane rotation, and horizontal translation. Fig. 6 shows the results over the 300 test images that were used in the previous subsection. We find that the model trained on segmentation masks is more robust than the other two models over all three geometric transformations, followed by the models trained on the object bounding boxes and the full image, in order. 3.4 QUALITATIVE INSPECTION OF OBJECTNET IMAGES AND ANNOTATIONS During the annotation of Object Net images, we came across the following observations: a) Some objects look very different when they are in motion (e.g., the fan in row 4 of Fig. 34 in Appx. D), or when they are shadowed or occluded by other objects (e.g., the hammer in Fig. 34 row 4), b) Some object instances differ a lot from the typical instances in the same class (e.g., the helmet in Fig. 34 row 5; the orange in Fig. 33 row 5), c) Some objects can be recognized only through reading their captions (e.g., the pet food container in Fig. 33 row 2), d) Some images have wrong labels (e.g., the pillow in Fig. 33 row 2; the skirt in Fig. 33 row 1; the tray in Fig. 34 row 2; See also Appx. E), e) Some objects are extremely difficult for humans (e.g., the tennis racket in Fig. 34 row 4; the shovel in Fig. 33 row 4; the tray in Fig. 33 row 1), f) In many images, objects are occluded by hands holding them (e.g., the sock and the shovel in Fig. 33 row 4), g) Some objects are hard to recognize in dim light (e.g., the printer in Fig. 33 row 2), and h) Some categories are often confused with other categories in the same set. Example sets include {bath towel, bed sheet, full sized towel, dishrag or hand towel}, {sandal, dress shoe (men), running shoe}, {t-shirt, dress, sweater, suit jacket, skirt}, {ruler, spatula, pen, match}, {padlock, combination lock}, and {envelope, letter}. The left panel in Fig. 7 shows four easy (highly confident correct predictions) and four hard (highly confident misclassifications) for Res Net-152 over six Object Net categories. In terms of the difficulty level, easy(difficult) objects for models appear easy(difficult) to humans too. Also, our qualitative inspection shows that Object Net includes a large number of objects that can be recognized only after a careful examination (the right panel in Fig. 7). More examples are given in Appx. C. Published as a conference paper at ICLR 2021 banana band aid ruler safety pin broom candle table lamp rock python stole nematode packet spatula nail paper towel letter opener marimba sundial solar dish washbashin syringe corckscrew bucket swab swab unicycle bobsled paper towel pencil box toilet tissue spindle Figure 7: Left) Four easiest (highest classification confidence; 1st rows) and four hardest (lowest confidence; 2nd rows) samples per category from the Object Net dataset for the Res Net-152 model (predictions shown in blue). The easy samples seem to be easier for humans (qualitatively), while the harder ones also seem to be harder. Right) A collection of challenging objects from the Object Net dataset for humans. Can you guess the category of the annotated objects in these images? Keys are: row 1: skirt, fan, desk lamp, safety pin, still camera, and spatula, row 2: vase, shovel, vacuum cleaner, printer, remote control, and pet food container, and row 3: sandal, vase, match, spatula, stuffed animal, and shovel. Appxs. C & D contain more examples. 4 TAKEAWAYS AND DISCUSSION Our investigation reveals that deep models perform significantly better when applied to isolated objects rather than the entire scene. The reason behind this is two-fold. First, there is less variability in single objects compared to scenes containing multiple objects. Second, deep models (used here and also in Object Net paper) have been trained on Image Net images which are less cluttered compared to the Object Net images. We anticipate that training models from scratch on large scale datasets that contain isolated objects will likely result in even higher accuracy. Assuming around 30% increase in performance (at best) on top of Barbu et al. s results using bounding boxes still leaves a large gap of at least 15% between Image Net and Object Net which means that Object Net is indeed much harder. It covers a wider range of variations than Image Net including object instances, viewpoints, rotations, occlusions, etc which pushes the limits of object recognition models. Hence, despite its limitations and biases, Object Net dataset remains a great platform to test deep models in realistic situations. We envision four research directions for the future work in this area. First, background subtraction is a promising mechanism and should be investigated further over large scale datasets (given the availability of high-resolution masks; e.g., MS COCO). We found that it improves robustness substantially over various types of image perturbations and attacks. Humans can discern the foreground object from the image background with high precision. This feat might be the key to robustness and hints towards an interplay and feedback loop between recognition and segmentation that is currently missing in CNNs. Second, measuring human performance on Object Net will provide a useful baseline for gauging model performance. Barbu et. al report an accuracy of around 95% when they asked subjects to mention the objects that are present in the scene. This task, however, is different from recognizing isolated objects similar to the regime that was considered here (i.e., akin to rapid scene categorization tasks; See Serre et al. (2007)). Besides, error patterns of models and humans (e.g., Borji & Itti (2014)), in addition to crude accuracy measures, will inform us about the differences in object recognition mechanisms between humans and machines. It could be that models work in a completely different fashion than the human visual system. Third, as discussed in Section 3.1, multi-label prediction accuracy is more appropriate for evaluating recognition models. Annotating all objects in Object Net images will thus provide an additional dimension to assess models. In this regard, we propose a new task where the goal is to recognize objects in their natural contexts. This task resembles (cropped) object recognition and object detection, but it is slightly different (i.e., the goal here is to recognize an object limited by a bounding box given all available information in the scene). This is essentially an argument against the recognition-detection dichotomy. Finally, it would be interesting to see how well the state of the art object detectors perform on the Object Net dataset (e.g., over overlapped classes between Object Net and MS COCO (Lin et al., 2014)). We expect a significant drop in detection performance since it is hard to recognize objects in this dataset. From a broader perspective, our study reinforces the idea that there is more to scene understanding then merely learning statistical correlations. In particular, background subtraction and visual context are crucial in robust recognition and demand further investigation in future studies. Published as a conference paper at ICLR 2021 Aharon Azulay and Yair Weiss. Why do deep convolutional networks generalize so poorly to small image transformations? Journal of Machine Learning Research, 20(184):1 25, 2019. Moshe Bar. Visual objects in context. Nature Reviews Neuroscience, 5(8):617, 2004. Andrei Barbu, David Mayo, Julian Alverio, William Luo, Christopher Wang, Dan Gutfreund, Josh Tenenbaum, and Boris Katz. Objectnet: A large-scale bias-controlled dataset for pushing the limits of object recognition models. In Advances in Neural Information Processing Systems, pp. 9448 9458, 2019. Sara Beery, Grant Van Horn, and Pietro Perona. Recognition in terra incognita. In Proceedings of the European Conference on Computer Vision (ECCV), pp. 456 473, 2018. Lucas Beyer, Olivier J H enaff, Alexander Kolesnikov, Xiaohua Zhai, and A aron van den Oord. Are we done with imagenet? ar Xiv preprint ar Xiv:2006.07159, 2020. Irving Biederman. Recognition-by-components: a theory of human image understanding. Psychological review, 94(2):115, 1987. Ali Borji. Objectnet dataset: Reanalysis and correction. ar Xiv preprint ar Xiv:2004.02042, 2020. Ali Borji and Laurent Itti. Human vs. computer in scene and object recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 113 120, 2014. Ali Borji, Saeed Izadi, and Laurent Itti. ilab-20m: A large-scale controlled object dataset to investigate deep learning. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2221 2230, 2016. Taco Cohen and Max Welling. Group equivariant convolutional networks. In International conference on machine learning, pp. 2990 2999, 2016. Alexander D Amour, Katherine Heller, Dan Moldovan, Ben Adlam, Babak Alipanahi, Alex Beutel, Christina Chen, Jonathan Deaton, Jacob Eisenstein, Matthew D Hoffman, et al. Underspecification presents challenges for credibility in modern machine learning. ar Xiv preprint ar Xiv:2011.03395, 2020. Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A largescale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pp. 248 255. Ieee, 2009. Santosh K Divvala, Derek Hoiem, James H Hays, Alexei A Efros, and Martial Hebert. An empirical study of context in object detection. In 2009 IEEE Conference on computer vision and Pattern Recognition, pp. 1271 1278. IEEE, 2009. Samuel Dodge and Lina Karam. A study and comparison of human and deep learning recognition performance under visual distortions. In 2017 26th international conference on computer communication and networks (ICCCN), pp. 1 7. IEEE, 2017. Logan Engstrom, Brandon Tran, Dimitris Tsipras, Ludwig Schmidt, and Aleksander Madry. Exploring the landscape of spatial robustness. In International Conference on Machine Learning, pp. 1802 1811, 2019. Alhussein Fawzi and Pascal Frossard. Manitest: Are classifiers really invariant? ar Xiv preprint ar Xiv:1507.06535, 2015. Carolina Galleguillos and Serge Belongie. Context based object categorization: A critical survey. Computer vision and image understanding, 114(6):712 722, 2010. Robert Geirhos, Patricia Rubisch, Claudio Michaelis, Matthias Bethge, Felix A Wichmann, and Wieland Brendel. Imagenet-trained cnns are biased towards texture; increasing shape bias improves accuracy and robustness. ar Xiv preprint ar Xiv:1811.12231, 2018a. Published as a conference paper at ICLR 2021 Robert Geirhos, Carlos RM Temme, Jonas Rauber, Heiko H Sch utt, Matthias Bethge, and Felix A Wichmann. Generalisation in humans and deep neural networks. In Advances in neural information processing systems, pp. 7538 7550, 2018b. Robert Geirhos, J orn-Henrik Jacobsen, Claudio Michaelis, Richard Zemel, Wieland Brendel, Matthias Bethge, and Felix A Wichmann. Shortcut learning in deep neural networks. Nature Machine Intelligence, 2(11):665 673, 2020. Ian J Goodfellow, Jonathon Shlens, and Christian Szegedy. Explaining and harnessing adversarial examples. ar Xiv preprint ar Xiv:1412.6572, 2014. Keren Gu, Brandon Yang, Jiquan Ngiam, Quoc Le, and Jonathon Shlens. Using videos to evaluate image model robustness. ar Xiv preprint ar Xiv:1904.10076, 2019. Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770 778, 2016. Geremy Heitz and Daphne Koller. Learning spatial context: Using stuff to find things. In European conference on computer vision, pp. 30 43. Springer, 2008. Dan Hendrycks and Thomas Dietterich. Benchmarking neural network robustness to common corruptions and perturbations. ar Xiv preprint ar Xiv:1903.12261, 2019. Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Sara Beery, et al. Wilds: A benchmark of in-the-wild distribution shifts. ar Xiv preprint ar Xiv:2012.07421, 2020. Nikolaus Kriegeskorte. Deep neural networks: a new framework for modeling biological vision and brain information processing. Annual review of vision science, 1:417 446, 2015. Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. 2009. Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems, pp. 1097 1105, 2012. Alina Kuznetsova, Hassan Rom, Neil Alldrin, Jasper Uijlings, Ivan Krasin, Jordi Pont-Tuset, Shahab Kamali, Stefan Popov, Matteo Malloci, Tom Duerig, et al. The open images dataset v4: Unified image classification, object detection, and visual relationship detection at scale. ar Xiv preprint ar Xiv:1811.00982, 2018. Yann Le Cun, L eon Bottou, Yoshua Bengio, and Patrick Haffner. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11):2278 2324, 1998. Yann Le Cun, Fu Jie Huang, and Leon Bottou. Learning methods for generic object recognition with invariance to pose and lighting. In Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004., volume 2, pp. II 104. IEEE, 2004. Han S Lee, Alex A Agarwal, and Junmo Kim. Why do deep neural networks still not recognize these images?: A qualitative analysis on failure cases of imagenet classification. ar Xiv preprint ar Xiv:1709.03439, 2017. Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In European conference on computer vision, pp. 740 755. Springer, 2014. Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu. Towards deep learning models resistant to adversarial attacks. ar Xiv preprint ar Xiv:1706.06083, 2017. Published as a conference paper at ICLR 2021 Tomasz Malisiewicz and Alexei A Efros. Improving spatial support for objects via multiple segmentations. 2007. Vaishnavh Nagarajan, Anders Andreassen, and Behnam Neyshabur. Understanding the failure modes of out-of-distribution generalization. ar Xiv preprint ar Xiv:2010.15775, 2020. Anh Nguyen, Jason Yosinski, and Jeff Clune. Deep neural networks are easily fooled: High confidence predictions for unrecognizable images. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 427 436, 2015. Andrew Rabinovich, Andrea Vedaldi, Carolina Galleguillos, Eric Wiewiora, and Serge J Belongie. Objects in context. In ICCV, volume 1, pp. 5. Citeseer, 2007. Benjamin Recht, Rebecca Roelofs, Ludwig Schmidt, and Vaishaal Shankar. Do imagenet classifiers generalize to imagenet? ar Xiv preprint ar Xiv:1902.10811, 2019. Amir Rosenfeld and Daphna Weinshall. Extracting foreground masks towards object recognition. In 2011 International Conference on Computer Vision, pp. 1371 1378. IEEE, 2011. Amir Rosenfeld, Richard Zemel, and John K Tsotsos. The elephant in the room. ar Xiv preprint ar Xiv:1808.03305, 2018. Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, et al. Imagenet large scale visual recognition challenge. International journal of computer vision, 115(3):211 252, 2015. Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad-cam: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision, pp. 618 626, 2017. Thomas Serre. Deep learning: the good, the bad, and the ugly. Annual Review of Vision Science, 5: 399 426, 2019. Thomas Serre, Aude Oliva, and Tomaso Poggio. A feedforward architecture accounts for rapid categorization. Proceedings of the national academy of sciences, 104(15):6424 6429, 2007. Vaishaal Shankar, Rebecca Roelofs, Horia Mania, Alex Fang, Benjamin Recht, and Ludwig Schmidt. Evaluating machine accuracy on imagenet. In International Conference on Machine Learning (ICML), 2020. Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. ar Xiv preprint ar Xiv:1409.1556, 2014. Bharat Singh and Larry S Davis. An analysis of scale invariance in object detection snip. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3578 3587, 2018. Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Ian Goodfellow, and Rob Fergus. Intriguing properties of neural networks. ar Xiv preprint ar Xiv:1312.6199, 2013. Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1 9, 2015. Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. Rethinking the inception architecture for computer vision. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2818 2826, 2016. Mingxing Tan, Bo Chen, Ruoming Pang, Vijay Vasudevan, Mark Sandler, Andrew Howard, and Quoc V Le. Mnasnet: Platform-aware neural architecture search for mobile. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2820 2828, 2019. Published as a conference paper at ICLR 2021 Rohan Taori, Achal Dave, Vaishaal Shankar, Nicholas Carlini, Benjamin Recht, and Ludwig Schmidt. Measuring robustness to natural distribution shifts in image classification. ar Xiv preprint ar Xiv:2007.00644, 2020. Antonio Torralba and Pawan Sinha. Statistical context priming for object detection. In Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001, volume 1, pp. 763 770. IEEE, 2001. Shimon Ullman, Liav Assif, Ethan Fetaya, and Daniel Harari. Atoms of recognition in human and computer vision. Proceedings of the National Academy of Sciences, 113(10):2744 2749, 2016. Anna Volokitin, Gemma Roig, and Tomaso A Poggio. Do deep neural networks suffer from crowding? In Advances in Neural Information Processing Systems, pp. 5628 5638, 2017. Kai Xiao, Logan Engstrom, Andrew Ilyas, and Aleksander Madry. Noise or signal: The role of image backgrounds in object recognition. ar Xiv preprint ar Xiv:2006.09994, 2020. Hongyang Zhang, Yaodong Yu, Jiantao Jiao, Eric P Xing, Laurent El Ghaoui, and Michael I Jordan. Theoretically principled trade-off between robustness and accuracy. ar Xiv preprint ar Xiv:1901.08573, 2019. Richard Zhang. Making convolutional networks shift-invariant again. ar Xiv preprint ar Xiv:1904.11486, 2019. Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba. Places: A 10 million image database for scene recognition. IEEE transactions on pattern analysis and machine intelligence, 40(6):1452 1464, 2017. Zhuotun Zhu, Lingxi Xie, and Alan L. Yuille. Object recognition with and without objects, 2016. Published as a conference paper at ICLR 2021 A FREQUENCY OF THE IMAGES PER CATEGORY Figure 8: Frequency of the images (one object label per image) over the 113 categories of Object Net overlapped with the Image Net. The right bar chart is the continuation of of the left one. Published as a conference paper at ICLR 2021 B MODEL ACCURACY PER CATEGORY USING BOXES VS. FULL IMAGE Figure 9: Performance of models on object bounding boxes (from left to right: Alex Net, VGG, and Goog Le Net). Published as a conference paper at ICLR 2021 Figure 10: Performance of models on object bounding boxes (from left to right: Res Net, Inception3, and MNASNet). Published as a conference paper at ICLR 2021 Figure 11: Performance of models on the entire image (from left to right: Alex Net, VGG, and Goog Le Net). Published as a conference paper at ICLR 2021 Figure 12: Performance of models on the entire image (from left to right: Res Net, Inception3, and MNASNet). Published as a conference paper at ICLR 2021 C EASIEST AND HARDEST OBJECTS FOR THE RESNET-152 MODEL (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 13: Correctly classified and misclassified examples from the Alarm clock class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 14: Correctly classified and misclassified examples from the Banana class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 15: Correctly classified and misclassified examples from the Band-Aid class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences. Only three benches were correctly classified. (b) Misclassified; highest confidences (c) Misclassified; lowest confidences Figure 16: Correctly classified and misclassified examples from the Bench class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences. Only two plates were correctly classified. (b) Misclassified; highest confidences (c) Misclassified; lowest confidences Figure 17: Correctly classified and misclassified examples from the Plate class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 18: Correctly classified and misclassified examples from the Broom class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 19: Correctly classified and misclassified examples from the Candle class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 20: Correctly classified and misclassified examples from the Fan class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 21: Correctly classified and misclassified examples from the Ruler class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 22: Correctly classified and misclassified examples from the Safety-pin class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 23: Correctly classified and misclassified examples from the Teapot class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 24: Correctly classified and misclassified examples from the TV class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 25: Correctly classified and misclassified examples from the Sock class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 26: Correctly classified and misclassified examples from the Sunglasses class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 27: Correctly classified and misclassified examples from the Tie class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 28: Correctly classified and misclassified examples from the Mug class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 29: Correctly classified and misclassified examples from the Hammer class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 30: Correctly classified and misclassified examples from the Bicycle class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 31: Correctly classified and misclassified examples from the Cellphone class by the Res Net model. Published as a conference paper at ICLR 2021 (a) Correctly classified; highest confidences (b) Correctly classified; lowest confidences (c) Misclassified; highest confidences (d) Misclassified; lowest confidences Figure 32: Correctly classified and misclassified examples from the Chair class by the Res Net model. Published as a conference paper at ICLR 2021 D SOME CHALLENGING EXAMPLES FOR HUMANS Figure 33: A selection of challenging objects that are hard to be recognized by humans. Can you guess the category of the annotated objects in these images? Keys are as follows: row 1: (skirt, skirt, desk lamp, safety pin, still camera, spatula, tray), row 2: (vase, pillow, sleeping bag, printer, remote control, pet food container, detergent), row 3: (vacuum cleaner, vase, vase, shovel, stuffed animal, sandal, sandal), row 4: (sock, shovel, shovel, skirt, skirt, match, spatula), row 5: (padlock, padlock, microwave, orange, printer, trash bin, tray) Published as a conference paper at ICLR 2021 Figure 34: A selection of challenging objects that are hard to be recognized by humans (continued from above). Can you guess the category of the annotated objects in these images? Keys are as follows: row 1: (remote control, ruler, full sized towel, ruler, remote control, remote control, remote control), row 2: (remote control, calendar, butter, bookend, ruler, tray, desk lamp), row 3: (envelope, envelope, drying rack for dishes, full sized towel, drying rack for dishes, drinking cup, desk lamp), row 4: (desk lamp, desk lamp, dress, tennis racket, fan, fan, hammer), row 5: (printer, toaster, printer, helmet, printer, printer, printer) Published as a conference paper at ICLR 2021 E ANNOTATION ISSUES IN OBJECT RECOGNITION DATASETS Figure 35: Top: Some examples from the Image Net dataset highlighting why multi-label annotations are necessary: (a) Multiple objects. Here, the image label is desk but screen, monitor, coffee mug and many more objects in the scene could count as correct labels, (b) Non-salient object. A scene where the target label picket fence is counter intuitive because it appears in the image background, while classes groom, bow-tie, suit, gown, and possibly hoopskirt are more prominently displayed in the foreground. c) Synonym or subset relationships: This image has Image Net label African elephant, but can be labeled tusker as well, because every African elephant with tusks is a tusker. d) Unclear images: This image is labeled lake shore, but could also be labeled seashore as there is not enough information in the scene to distinguish the water body between a lake or sea. Image compiled from (Shankar et al., 2020). Bottom: Some annotation problems in the Image Net dataset. Example images along with their ground truth (GT) labels (red) and predicted classes (PC) by a model (blue) are shown. Top-left) Similar labels, Top-right) non-salient GT, Bottom-left) challenging images, and Bottom-right) incorrect GT. Please see Lee et al. (2017) for details. Published as a conference paper at ICLR 2021 F ANALYSING MODEL ROBUSTNESS OVER NATURALLY DISTORTED IMAGES Figure 36: Sample images alongside their corruptions at 3 severity levels. Published as a conference paper at ICLR 2021 Figure 37: Top-1 (dashed lines) and Top-5 (solid lines) accuracy of models over 14 natural image distortions at 3 severity levels (using object bounding box). Published as a conference paper at ICLR 2021 Figure 38: Top-1 (dashed lines) and Top-5 (solid lines) accuracy of models over 14 natural image distortions at 3 severity levels (using full image). Published as a conference paper at ICLR 2021 G ADVERSARIAL DEFENSE USING FOREGROUND DETECTION ON MNIST AND FASHION MNIST Figure 39: Top-left) Model, trained on clean MNIST data, is attacked by FGSM, Top-right) Foreground detection over a model that has been trained on clean data. Bottom-left) Model, trained on noisy MNIST data, is attacked by FGSM, Bottom-right) Foreground detection of a model that has been trained on noisy data. Noisy data is created by overlaying an object over white noise i.e., noise (1mask) + object. We find that background subtraction together with edge detection improves robustness. Published as a conference paper at ICLR 2021 Figure 40: Top-left) Model, trained on clean Fashion MNIST data, is attacked by FGSM, Top-right) Foreground detection over a model that has been trained on clean data. Bottom-left) Model, trained on noisy Fashion MNIST data, is attacked by FGSM, Bottom-right) Foreground detection of a model that has been trained on noisy data. Noisy data is created by overlaying an object over white noise i.e., noise (1mask) + object. We find that background subtraction together with edge detection improves robustness. Published as a conference paper at ICLR 2021 H STATISTICS OF OBJECTNET DATASET Figure 41: Top: Distribution of object scale in Object Net dataset (113 annotated classes), Bottom: Distribution object scale in ILSVRC2012-2014 single-object localization (dark green) and PASCAL VOC 2012 (light blue) validation sets. Object scale is fraction of image area occupied by an average object instance. Please see Fig. 16 in Russakovsky et al. (2015) . Figure 42: Distribution of object location in Object Net (113 annotated classes).