Increasing CNN Robustness to Occlusions by Reducing Filter Support

Osherov, Elad; Lindenbaum, Michael

doi:10.1109/iccv.2017.67

Cited by 39 publications

(24 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The authors of [19] already evaluated CNNs trained on occlusions in the context of 2D object detection and recognition and proposed modifying training to penalize large spatial filters support. This yields better performance; however, this does not fully cancel out the influence of occlusions.…”

Section: Related Workmentioning

confidence: 99%

Making Deep Heatmaps Robust to Partial Occlusions for 3D Object Pose Estimation

Oberweger

Rad

Lepetit

2018

Computer Vision – ECCV 2018

208

191

View full text Add to dashboard Cite

We introduce a novel method for robust and accurate 3D object pose estimation from a single color image under large occlusions. Following recent approaches, we first predict the 2D projections of 3D points related to the target object and then compute the 3D pose from these correspondences using a geometric method. Unfortunately, as the results of our experiments show, predicting these 2D projections using a regular CNN or a Convolutional Pose Machine is highly sensitive to partial occlusions, even when these methods are trained with partially occluded examples. Our solution is to predict heatmaps from multiple small patches independently and to accumulate the results to obtain accurate and robust predictions. Training subsequently becomes challenging because patches with similar appearances but different positions on the object correspond to different heatmaps. However, we provide a simple yet effective solution to deal with such ambiguities. We show that our approach outperforms existing methods on two challenging datasets: The Occluded LineMOD dataset and the YCB-Video dataset, both exhibiting cluttered scenes with highly occluded objects.

show abstract

Section: Related Workmentioning

confidence: 99%

Making Deep Heatmaps Robust to Partial Occlusions for 3D Object Pose Estimation

Oberweger

Rad

Lepetit

2018

Computer Vision – ECCV 2018

208

191

View full text Add to dashboard Cite

show abstract

“…Such models perform poorly for occluded images, which is the case for real life scenarios of fruit classification. In a CNN while using convolutional layer, features extracted from convolutional layer are spatially local [OL17] and due to use of only convolutional layers for occluded images some important features defining the object in the image could be lost, hence, failing to classify occluded images of fruits & vegetables properly during testing. Therefore, using a mixture of fully connected layers along with convolutional layers could resolve issues related to occlusion in images.…”

Section: Pre-trained Network and Transfer Learningmentioning

confidence: 99%

FruitVegCNN: Power- and Memory-Efficient Classification of Fruits & Vegetables Using CNN in Mobile MPSoC

Dey¹,

Saha²,

Singh³

et al. 2020

Preprint

View full text Add to dashboard Cite

<div><div><div><p>Fruit and vegetable classification using Convolutional Neural Networks (CNNs) has become a popular application in the agricultural industry, however, to the best of our knowledge no previously recorded study has designed and evaluated such an application on a mobile platform. In this paper, we propose a power-efficient CNN model, FruitVegCNN, to perform classification of fruits and vegetables in a mobile multi-processor system-on-a-chip (MPSoC). We also evaluated the efficacy of FruitVegCNN compared to popular state-of-the-art CNN models in real mobile plat- forms (Huawei P20 Lite and Samsung Galaxy Note 9) and experimental results show the efficacy and power efficiency of our proposed CNN architecture.</p></div></div></div>

show abstract

“…To solve the occlusion-related problems, the strength of the features is being usually improved through designated regularization, i.e., L1-norm [11]. This notion is considered to be appropriate for non-facial occlusions, e.g., Osherov and Lindenbaum suggested to jointly work out for the re-weight L1 regularization to tackle with arbitrary occlusions in the end-to-end framework [12].…”

Section: Holistic-based Approachesmentioning

confidence: 99%

Facial Expression Recognition of Nonlinear Facial Variations Using Deep Locality De-Expression Residue Learning in the Wild

et al. 2019

View full text Add to dashboard Cite

Automatic facial expression recognition is an emerging field. Moreover, the interest has been increased with the transition from laboratory-controlled conditions to in the wild scenarios. Most of the research has been done over nonoccluded faces under the constrained environment, while automatic facial expression is less understood/implemented for partial occlusion in the real world conditions. Apart from that, our research aims to tackle the issues of overfitting (caused by the shortage of adequate training data) and to alleviate the expression-unrelated/intraclass/nonlinear facial variations, such as head pose estimation, eye gaze estimation, intensity and microexpressions. In our research, we control the magnitude of each Action Unit (AU) and combine several of the Action Unit combinations to leverage learning from the generative and discriminative representations for automatic FER. We have also addressed the problem of diversification of expressions from lab controlled to real-world scenarios from our cross-database study and proposed a model for enhancement of the discriminative power of deep features while increasing the interclass scatters, by preserving the locality closeness. Furthermore, facial expression consists of an expressive component as well as neutral component, so we proposed a generative model which is capable of generating neutral expression from an input image using cGAN. The expressive component is filtered and passed to the intermediate layers and the process is called De-expression Residue Learning. The residue in the intermediate/middle layers is very important for learning through expressive components. Finally, we validate the effectiveness of our method (DLP-DeRL) through qualitative and quantitative experimental results using four databases. Our method is more accurate and robust, and outperforms all the existing methods (hand crafted features and deep learning) while dealing the images in the wild.

show abstract

Increasing CNN Robustness to Occlusions by Reducing Filter Support

Cited by 39 publications

References 13 publications

Making Deep Heatmaps Robust to Partial Occlusions for 3D Object Pose Estimation

Making Deep Heatmaps Robust to Partial Occlusions for 3D Object Pose Estimation

FruitVegCNN: Power- and Memory-Efficient Classification of Fruits & Vegetables Using CNN in Mobile MPSoC

Facial Expression Recognition of Nonlinear Facial Variations Using Deep Locality De-Expression Residue Learning in the Wild

Contact Info

Product

Resources

About