Pub Date : 2022-09-09DOI: 10.48550/arXiv.2209.04284
Zhewen Zhang, Fuliang Wu, Yuming Qiu, Jingdong Liang, Shuiwang Li
With more and more large-scale datasets available for training, visual tracking has made great progress in recent years. However, current research in the field mainly focuses on tracking generic objects. In this paper, we present TSFMO, a benchmark for textbf{T}racking textbf{S}mall and textbf{F}ast textbf{M}oving textbf{O}bjects. This benchmark aims to encourage research in developing novel and accurate methods for this challenging task particularly. TSFMO consists of 250 sequences with about 50k frames in total. Each frame in these sequences is carefully and manually annotated with a bounding box. To the best of our knowledge, TSFMO is the first benchmark dedicated to tracking small and fast moving objects, especially connected to sports. To understand how existing methods perform and to provide comparison for future research on TSFMO, we extensively evaluate 20 state-of-the-art trackers on the benchmark. The evaluation results exhibit that more effort are required to improve tracking small and fast moving objects. Moreover, to encourage future research, we proposed a novel tracker S-KeepTrack which surpasses all 20 evaluated approaches. By releasing TSFMO, we expect to facilitate future researches and applications of tracking small and fast moving objects. The TSFMO and evaluation results as well as S-KeepTrack are available at url{https://github.com/CodeOfGithub/S-KeepTrack}.
{"title":"Tracking Small and Fast Moving Objects: A Benchmark","authors":"Zhewen Zhang, Fuliang Wu, Yuming Qiu, Jingdong Liang, Shuiwang Li","doi":"10.48550/arXiv.2209.04284","DOIUrl":"https://doi.org/10.48550/arXiv.2209.04284","url":null,"abstract":"With more and more large-scale datasets available for training, visual tracking has made great progress in recent years. However, current research in the field mainly focuses on tracking generic objects. In this paper, we present TSFMO, a benchmark for textbf{T}racking textbf{S}mall and textbf{F}ast textbf{M}oving textbf{O}bjects. This benchmark aims to encourage research in developing novel and accurate methods for this challenging task particularly. TSFMO consists of 250 sequences with about 50k frames in total. Each frame in these sequences is carefully and manually annotated with a bounding box. To the best of our knowledge, TSFMO is the first benchmark dedicated to tracking small and fast moving objects, especially connected to sports. To understand how existing methods perform and to provide comparison for future research on TSFMO, we extensively evaluate 20 state-of-the-art trackers on the benchmark. The evaluation results exhibit that more effort are required to improve tracking small and fast moving objects. Moreover, to encourage future research, we proposed a novel tracker S-KeepTrack which surpasses all 20 evaluated approaches. By releasing TSFMO, we expect to facilitate future researches and applications of tracking small and fast moving objects. The TSFMO and evaluation results as well as S-KeepTrack are available at url{https://github.com/CodeOfGithub/S-KeepTrack}.","PeriodicalId":87238,"journal":{"name":"Computer vision - ACCV ... : ... Asian Conference on Computer Vision : proceedings. Asian Conference on Computer Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75253036","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In object detection, data amount and cost are a trade-off, and collecting a large amount of data in a specific domain is labor intensive. Therefore, existing large-scale datasets are used for pre-training. However, conventional transfer learning and domain adaptation cannot bridge the domain gap when the target domain differs significantly from the source domain. We propose a data synthesis method that can solve the large domain gap problem. In this method, a part of the target image is pasted onto the source image, and the position of the pasted region is aligned by utilizing the information of the object bounding box. In addition, we introduce adversarial learning to discriminate whether the original or the pasted regions. The proposed method trains on a large number of source images and a few target domain images. The proposed method achieves higher accuracy than conventional methods in a very different domain problem setting, where RGB images are the source domain, and thermal infrared images are the target domain. Similarly, the proposed method achieves higher accuracy in the cases of simulation images to real images.
{"title":"Few-shot Adaptive Object Detection with Cross-Domain CutMix","authors":"Yuzuru Nakamura, Yasunori Ishii, Yuki Maruyama, Takayoshi Yamashita","doi":"10.48550/arXiv.2208.14586","DOIUrl":"https://doi.org/10.48550/arXiv.2208.14586","url":null,"abstract":"In object detection, data amount and cost are a trade-off, and collecting a large amount of data in a specific domain is labor intensive. Therefore, existing large-scale datasets are used for pre-training. However, conventional transfer learning and domain adaptation cannot bridge the domain gap when the target domain differs significantly from the source domain. We propose a data synthesis method that can solve the large domain gap problem. In this method, a part of the target image is pasted onto the source image, and the position of the pasted region is aligned by utilizing the information of the object bounding box. In addition, we introduce adversarial learning to discriminate whether the original or the pasted regions. The proposed method trains on a large number of source images and a few target domain images. The proposed method achieves higher accuracy than conventional methods in a very different domain problem setting, where RGB images are the source domain, and thermal infrared images are the target domain. Similarly, the proposed method achieves higher accuracy in the cases of simulation images to real images.","PeriodicalId":87238,"journal":{"name":"Computer vision - ACCV ... : ... Asian Conference on Computer Vision : proceedings. Asian Conference on Computer Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79770867","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-07DOI: 10.48550/arXiv.2208.03660
Yujiao Shi, Xin Yu, Shanhe Wang, Hongdong Li
This paper tackles the problem of Cross-view Video-based camera Localization (CVL). The task is to localize a query camera by leveraging information from its past observations, i.e., a continuous sequence of images observed at previous time stamps, and matching them to a large overhead-view satellite image. The critical challenge of this task is to learn a powerful global feature descriptor for the sequential ground-view images while considering its domain alignment with reference satellite images. For this purpose, we introduce CVLNet, which first projects the sequential ground-view images into an overhead view by exploring the ground-and-overhead geometric correspondences and then leverages the photo consistency among the projected images to form a global representation. In this way, the cross-view domain differences are bridged. Since the reference satellite images are usually pre-cropped and regularly sampled, there is always a misalignment between the query camera location and its matching satellite image center. Motivated by this, we propose estimating the query camera's relative displacement to a satellite image before similarity matching. In this displacement estimation process, we also consider the uncertainty of the camera location. For example, a camera is unlikely to be on top of trees. To evaluate the performance of the proposed method, we collect satellite images from Google Map for the KITTI dataset and construct a new cross-view video-based localization benchmark dataset, KITTI-CVL. Extensive experiments have demonstrated the effectiveness of video-based localization over single image-based localization and the superiority of each proposed module over other alternatives.
{"title":"CVLNet: Cross-View Semantic Correspondence Learning for Video-based Camera Localization","authors":"Yujiao Shi, Xin Yu, Shanhe Wang, Hongdong Li","doi":"10.48550/arXiv.2208.03660","DOIUrl":"https://doi.org/10.48550/arXiv.2208.03660","url":null,"abstract":"This paper tackles the problem of Cross-view Video-based camera Localization (CVL). The task is to localize a query camera by leveraging information from its past observations, i.e., a continuous sequence of images observed at previous time stamps, and matching them to a large overhead-view satellite image. The critical challenge of this task is to learn a powerful global feature descriptor for the sequential ground-view images while considering its domain alignment with reference satellite images. For this purpose, we introduce CVLNet, which first projects the sequential ground-view images into an overhead view by exploring the ground-and-overhead geometric correspondences and then leverages the photo consistency among the projected images to form a global representation. In this way, the cross-view domain differences are bridged. Since the reference satellite images are usually pre-cropped and regularly sampled, there is always a misalignment between the query camera location and its matching satellite image center. Motivated by this, we propose estimating the query camera's relative displacement to a satellite image before similarity matching. In this displacement estimation process, we also consider the uncertainty of the camera location. For example, a camera is unlikely to be on top of trees. To evaluate the performance of the proposed method, we collect satellite images from Google Map for the KITTI dataset and construct a new cross-view video-based localization benchmark dataset, KITTI-CVL. Extensive experiments have demonstrated the effectiveness of video-based localization over single image-based localization and the superiority of each proposed module over other alternatives.","PeriodicalId":87238,"journal":{"name":"Computer vision - ACCV ... : ... Asian Conference on Computer Vision : proceedings. Asian Conference on Computer Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75838313","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-26DOI: 10.48550/arXiv.2207.12984
Jawad Tayyub, M. Sarmad, Nicolas Schonborn
Explaining decisions made by deep neural networks is a rapidly advancing research topic. In recent years, several approaches have attempted to provide visual explanations of decisions made by neural networks designed for structured 2D image input data. In this paper, we propose a novel approach to generate coarse visual explanations of networks designed to classify unstructured 3D data, namely point clouds. Our method uses gradients flowing back to the final feature map layers and maps these values as contributions of the corresponding points in the input point cloud. Due to dimensionality disagreement and lack of spatial consistency between input points and final feature maps, our approach combines gradients with points dropping to compute explanations of different parts of the point cloud iteratively. The generality of our approach is tested on various point cloud classification networks, including 'single object' networks PointNet, PointNet++, DGCNN, and a 'scene' network VoteNet. Our method generates symmetric explanation maps that highlight important regions and provide insight into the decision-making process of network architectures. We perform an exhaustive evaluation of trust and interpretability of our explanation method against comparative approaches using quantitative, quantitative and human studies. All our code is implemented in PyTorch and will be made publicly available.
{"title":"Explaining Deep Neural Networks for Point Clouds using Gradient-based Visualisations","authors":"Jawad Tayyub, M. Sarmad, Nicolas Schonborn","doi":"10.48550/arXiv.2207.12984","DOIUrl":"https://doi.org/10.48550/arXiv.2207.12984","url":null,"abstract":"Explaining decisions made by deep neural networks is a rapidly advancing research topic. In recent years, several approaches have attempted to provide visual explanations of decisions made by neural networks designed for structured 2D image input data. In this paper, we propose a novel approach to generate coarse visual explanations of networks designed to classify unstructured 3D data, namely point clouds. Our method uses gradients flowing back to the final feature map layers and maps these values as contributions of the corresponding points in the input point cloud. Due to dimensionality disagreement and lack of spatial consistency between input points and final feature maps, our approach combines gradients with points dropping to compute explanations of different parts of the point cloud iteratively. The generality of our approach is tested on various point cloud classification networks, including 'single object' networks PointNet, PointNet++, DGCNN, and a 'scene' network VoteNet. Our method generates symmetric explanation maps that highlight important regions and provide insight into the decision-making process of network architectures. We perform an exhaustive evaluation of trust and interpretability of our explanation method against comparative approaches using quantitative, quantitative and human studies. All our code is implemented in PyTorch and will be made publicly available.","PeriodicalId":87238,"journal":{"name":"Computer vision - ACCV ... : ... Asian Conference on Computer Vision : proceedings. Asian Conference on Computer Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90711816","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-25DOI: 10.48550/arXiv.2207.12392
M. Sultana, Muzammal Naseer, Muhammad Haris Khan, Salman Khan, F. Khan
In the recent past, several domain generalization (DG) methods have been proposed, showing encouraging performance, however, almost all of them build on convolutional neural networks (CNNs). There is little to no progress on studying the DG performance of vision transformers (ViTs), which are challenging the supremacy of CNNs on standard benchmarks, often built on i.i.d assumption. This renders the real-world deployment of ViTs doubtful. In this paper, we attempt to explore ViTs towards addressing the DG problem. Similar to CNNs, ViTs also struggle in out-of-distribution scenarios and the main culprit is overfitting to source domains. Inspired by the modular architecture of ViTs, we propose a simple DG approach for ViTs, coined as self-distillation for ViTs. It reduces the overfitting of source domains by easing the learning of input-output mapping problem through curating non-zero entropy supervisory signals for intermediate transformer blocks. Further, it does not introduce any new parameters and can be seamlessly plugged into the modular composition of different ViTs. We empirically demonstrate notable performance gains with different DG baselines and various ViT backbones in five challenging datasets. Moreover, we report favorable performance against recent state-of-the-art DG methods. Our code along with pre-trained models are publicly available at: https://github.com/maryam089/SDViT.
{"title":"Self-Distilled Vision Transformer for Domain Generalization","authors":"M. Sultana, Muzammal Naseer, Muhammad Haris Khan, Salman Khan, F. Khan","doi":"10.48550/arXiv.2207.12392","DOIUrl":"https://doi.org/10.48550/arXiv.2207.12392","url":null,"abstract":"In the recent past, several domain generalization (DG) methods have been proposed, showing encouraging performance, however, almost all of them build on convolutional neural networks (CNNs). There is little to no progress on studying the DG performance of vision transformers (ViTs), which are challenging the supremacy of CNNs on standard benchmarks, often built on i.i.d assumption. This renders the real-world deployment of ViTs doubtful. In this paper, we attempt to explore ViTs towards addressing the DG problem. Similar to CNNs, ViTs also struggle in out-of-distribution scenarios and the main culprit is overfitting to source domains. Inspired by the modular architecture of ViTs, we propose a simple DG approach for ViTs, coined as self-distillation for ViTs. It reduces the overfitting of source domains by easing the learning of input-output mapping problem through curating non-zero entropy supervisory signals for intermediate transformer blocks. Further, it does not introduce any new parameters and can be seamlessly plugged into the modular composition of different ViTs. We empirically demonstrate notable performance gains with different DG baselines and various ViT backbones in five challenging datasets. Moreover, we report favorable performance against recent state-of-the-art DG methods. Our code along with pre-trained models are publicly available at: https://github.com/maryam089/SDViT.","PeriodicalId":87238,"journal":{"name":"Computer vision - ACCV ... : ... Asian Conference on Computer Vision : proceedings. Asian Conference on Computer Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73853866","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-20DOI: 10.48550/arXiv.2207.10075
Chuhan Zhang, Ankush Gupta, Andrew Zisserman
The objective of this work is to learn an object-centric video representation, with the aim of improving transferability to novel tasks, i.e., tasks different from the pre-training task of action classification. To this end, we introduce a new object-centric video recognition model based on a transformer architecture. The model learns a set of object-centric summary vectors for the video, and uses these vectors to fuse the visual and spatio-temporal trajectory 'modalities' of the video clip. We also introduce a novel trajectory contrast loss to further enhance objectness in these summary vectors. With experiments on four datasets -- SomethingSomething-V2, SomethingElse, Action Genome and EpicKitchens -- we show that the object-centric model outperforms prior video representations (both object-agnostic and object-aware), when: (1) classifying actions on unseen objects and unseen environments; (2) low-shot learning of novel classes; (3) linear probe to other downstream tasks; as well as (4) for standard action classification.
{"title":"Is an Object-Centric Video Representation Beneficial for Transfer?","authors":"Chuhan Zhang, Ankush Gupta, Andrew Zisserman","doi":"10.48550/arXiv.2207.10075","DOIUrl":"https://doi.org/10.48550/arXiv.2207.10075","url":null,"abstract":"The objective of this work is to learn an object-centric video representation, with the aim of improving transferability to novel tasks, i.e., tasks different from the pre-training task of action classification. To this end, we introduce a new object-centric video recognition model based on a transformer architecture. The model learns a set of object-centric summary vectors for the video, and uses these vectors to fuse the visual and spatio-temporal trajectory 'modalities' of the video clip. We also introduce a novel trajectory contrast loss to further enhance objectness in these summary vectors. With experiments on four datasets -- SomethingSomething-V2, SomethingElse, Action Genome and EpicKitchens -- we show that the object-centric model outperforms prior video representations (both object-agnostic and object-aware), when: (1) classifying actions on unseen objects and unseen environments; (2) low-shot learning of novel classes; (3) linear probe to other downstream tasks; as well as (4) for standard action classification.","PeriodicalId":87238,"journal":{"name":"Computer vision - ACCV ... : ... Asian Conference on Computer Vision : proceedings. Asian Conference on Computer Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-07-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84771454","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-12DOI: 10.48550/arXiv.2207.05605
Tian Ye, Sixiang Chen, Yun Liu, Y. Ye, Erkang Chen
In winter scenes, the degradation of images taken under snow can be pretty complex, where the spatial distribution of snowy degradation is varied from image to image. Recent methods adopt deep neural networks to directly recover clean scenes from snowy images. However, due to the paradox caused by the variation of complex snowy degradation, achieving reliable High-Definition image desnowing performance in real time is a considerable challenge. We develop a novel Efficient Pyramid Network with asymmetrical encoder-decoder architecture for real-time HD image desnowing. The general idea of our proposed network is to utilize the multi-scale feature flow fully and implicitly mine clean cues from features. Compared with previous state-of-the-art desnowing methods, our approach achieves a better complexity-performance trade-off and effectively handles the processing difficulties of HD and Ultra-HD images. The extensive experiments on three large-scale image desnowing datasets demonstrate that our method surpasses all state-of-the-art approaches by a large margin both quantitatively and qualitatively, boosting the PSNR metric from 31.76 dB to 34.10 dB on the CSD test dataset and from 28.29 dB to 30.87 dB on the SRRS test dataset.
{"title":"Towards Real-time High-Definition Image Snow Removal: Efficient Pyramid Network with Asymmetrical Encoder-decoder Architecture","authors":"Tian Ye, Sixiang Chen, Yun Liu, Y. Ye, Erkang Chen","doi":"10.48550/arXiv.2207.05605","DOIUrl":"https://doi.org/10.48550/arXiv.2207.05605","url":null,"abstract":"In winter scenes, the degradation of images taken under snow can be pretty complex, where the spatial distribution of snowy degradation is varied from image to image. Recent methods adopt deep neural networks to directly recover clean scenes from snowy images. However, due to the paradox caused by the variation of complex snowy degradation, achieving reliable High-Definition image desnowing performance in real time is a considerable challenge. We develop a novel Efficient Pyramid Network with asymmetrical encoder-decoder architecture for real-time HD image desnowing. The general idea of our proposed network is to utilize the multi-scale feature flow fully and implicitly mine clean cues from features. Compared with previous state-of-the-art desnowing methods, our approach achieves a better complexity-performance trade-off and effectively handles the processing difficulties of HD and Ultra-HD images. The extensive experiments on three large-scale image desnowing datasets demonstrate that our method surpasses all state-of-the-art approaches by a large margin both quantitatively and qualitatively, boosting the PSNR metric from 31.76 dB to 34.10 dB on the CSD test dataset and from 28.29 dB to 30.87 dB on the SRRS test dataset.","PeriodicalId":87238,"journal":{"name":"Computer vision - ACCV ... : ... Asian Conference on Computer Vision : proceedings. Asian Conference on Computer Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91095977","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-08DOI: 10.48550/arXiv.2207.04132
Hannah Kim, Shuzhi Yu, Shuaihang Yuan, Carlo Tomasi
We propose TAIN (Transformers and Attention for video INterpolation), a residual neural network for video interpolation, which aims to interpolate an intermediate frame given two consecutive image frames around it. We first present a novel vision transformer module, named Cross Similarity (CS), to globally aggregate input image features with similar appearance as those of the predicted interpolated frame. These CS features are then used to refine the interpolated prediction. To account for occlusions in the CS features, we propose an Image Attention (IA) module to allow the network to focus on CS features from one frame over those of the other. TAIN outperforms existing methods that do not require flow estimation and performs comparably to flow-based methods while being computationally efficient in terms of inference time on Vimeo90k, UCF101, and SNU-FILM benchmarks.
我们提出了TAIN (Transformers and Attention for video INterpolation),这是一个用于视频插值的残差神经网络,其目的是在给定两个连续图像帧的情况下插值中间帧。我们首先提出了一种新的视觉转换模块,称为交叉相似度(CS),用于全局聚合与预测插值帧相似的输入图像特征。然后使用这些CS特征来改进插值预测。为了考虑CS特征中的遮挡,我们提出了一个图像注意(IA)模块,允许网络从一帧中关注CS特征,而不是其他帧。在Vimeo90k、UCF101和SNU-FILM基准测试中,TAIN的性能优于不需要流量估计的现有方法,与基于流量的方法相当,同时在推理时间方面计算效率很高。
{"title":"Cross-Attention Transformer for Video Interpolation","authors":"Hannah Kim, Shuzhi Yu, Shuaihang Yuan, Carlo Tomasi","doi":"10.48550/arXiv.2207.04132","DOIUrl":"https://doi.org/10.48550/arXiv.2207.04132","url":null,"abstract":"We propose TAIN (Transformers and Attention for video INterpolation), a residual neural network for video interpolation, which aims to interpolate an intermediate frame given two consecutive image frames around it. We first present a novel vision transformer module, named Cross Similarity (CS), to globally aggregate input image features with similar appearance as those of the predicted interpolated frame. These CS features are then used to refine the interpolated prediction. To account for occlusions in the CS features, we propose an Image Attention (IA) module to allow the network to focus on CS features from one frame over those of the other. TAIN outperforms existing methods that do not require flow estimation and performs comparably to flow-based methods while being computationally efficient in terms of inference time on Vimeo90k, UCF101, and SNU-FILM benchmarks.","PeriodicalId":87238,"journal":{"name":"Computer vision - ACCV ... : ... Asian Conference on Computer Vision : proceedings. Asian Conference on Computer Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88659260","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Channel pruning is widely used to reduce the complexity of deep network models. Recent pruning methods usually identify which parts of the network to discard by proposing a channel importance criterion. However, recent studies have shown that these criteria do not work well in all conditions. In this paper, we propose a novel Feature Shift Minimization (FSM) method to compress CNN models, which evaluates the feature shift by converging the information of both features and filters. Specifically, we first investigate the compression efficiency with some prevalent methods in different layer-depths and then propose the feature shift concept. Then, we introduce an approximation method to estimate the magnitude of the feature shift, since it is difficult to compute it directly. Besides, we present a distribution-optimization algorithm to compensate for the accuracy loss and improve the network compression efficiency. The proposed method yields state-of-the-art performance on various benchmark networks and datasets, verified by extensive experiments. Our codes are available at: https://github.com/lscgx/FSM.
{"title":"Network Pruning via Feature Shift Minimization","authors":"Y. Duan, Xiaofang Hu, Yue Zhou, Peng He, Qi Liu, Shukai Duan","doi":"10.48550/arXiv.2207.02632","DOIUrl":"https://doi.org/10.48550/arXiv.2207.02632","url":null,"abstract":"Channel pruning is widely used to reduce the complexity of deep network models. Recent pruning methods usually identify which parts of the network to discard by proposing a channel importance criterion. However, recent studies have shown that these criteria do not work well in all conditions. In this paper, we propose a novel Feature Shift Minimization (FSM) method to compress CNN models, which evaluates the feature shift by converging the information of both features and filters. Specifically, we first investigate the compression efficiency with some prevalent methods in different layer-depths and then propose the feature shift concept. Then, we introduce an approximation method to estimate the magnitude of the feature shift, since it is difficult to compute it directly. Besides, we present a distribution-optimization algorithm to compensate for the accuracy loss and improve the network compression efficiency. The proposed method yields state-of-the-art performance on various benchmark networks and datasets, verified by extensive experiments. Our codes are available at: https://github.com/lscgx/FSM.","PeriodicalId":87238,"journal":{"name":"Computer vision - ACCV ... : ... Asian Conference on Computer Vision : proceedings. Asian Conference on Computer Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83081464","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}