Pub Date : 2021-10-01DOI: 10.1109/ICCV48922.2021.00731
Jae-Hun Lee, ChanYoung Kim, S. Sull
Most supervised image segmentation methods require delicate and time-consuming pixel-level labeling of building or objects, especially for small objects. In this paper, we present a weakly supervised segmentation network for aerial/satellite images, separately considering small and large objects. First, we propose a simple point labeling method for small objects, while large objects are fully labeled. Then, we present a segmentation network trained with a small object mask to separate small and large objects in the loss function. During training, we employ a memory bank to cope with the limited number of point labels. Experiments results with three public datasets demonstrate the feasibility of our approach.
{"title":"Weakly Supervised Segmentation of Small Buildings with Point Labels","authors":"Jae-Hun Lee, ChanYoung Kim, S. Sull","doi":"10.1109/ICCV48922.2021.00731","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.00731","url":null,"abstract":"Most supervised image segmentation methods require delicate and time-consuming pixel-level labeling of building or objects, especially for small objects. In this paper, we present a weakly supervised segmentation network for aerial/satellite images, separately considering small and large objects. First, we propose a simple point labeling method for small objects, while large objects are fully labeled. Then, we present a segmentation network trained with a small object mask to separate small and large objects in the loss function. During training, we employ a memory bank to cope with the limited number of point labels. Experiments results with three public datasets demonstrate the feasibility of our approach.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"17 1","pages":"7386-7395"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90498586","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 : 2021-10-01DOI: 10.1109/ICCV48922.2021.00612
Taekyung Kim, Jaehoon Choi, Seokeon Choi, Dongki Jung, Changick Kim
While learning-based multi-view stereo (MVS) methods have recently shown successful performances in quality and efficiency, limited MVS data hampers generalization to unseen environments. A simple solution is to generate various large-scale MVS datasets, but generating dense ground truth for 3D structure requires a huge amount of time and resources. On the other hand, if the reliance on dense ground truth is relaxed, MVS systems will generalize more smoothly to new environments. To this end, we first introduce a novel semi-supervised multi-view stereo framework called a Sparse Ground truth-based MVS Network (SGT-MVSNet) that can reliably reconstruct the 3D structures even with a few ground truth 3D points. Our strategy is to divide the accurate and erroneous regions and individually conquer them based on our observation that a probability map can separate these regions. We propose a self-supervision loss called the 3D Point Consistency Loss to enhance the 3D reconstruction performance, which forces the 3D points back-projected from the corresponding pixels by the predicted depth values to meet at the same 3D co-ordinates. Finally, we propagate these improved depth pre-dictions toward edges and occlusions by the Coarse-to-fine Reliable Depth Propagation module. We generate the spare ground truth of the DTU dataset for evaluation and extensive experiments verify that our SGT-MVSNet outperforms the state-of-the-art MVS methods on the sparse ground truth setting. Moreover, our method shows comparable reconstruction results to the supervised MVS methods though we only used tens and hundreds of ground truth 3D points.
{"title":"Just a Few Points are All You Need for Multi-view Stereo: A Novel Semi-supervised Learning Method for Multi-view Stereo","authors":"Taekyung Kim, Jaehoon Choi, Seokeon Choi, Dongki Jung, Changick Kim","doi":"10.1109/ICCV48922.2021.00612","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.00612","url":null,"abstract":"While learning-based multi-view stereo (MVS) methods have recently shown successful performances in quality and efficiency, limited MVS data hampers generalization to unseen environments. A simple solution is to generate various large-scale MVS datasets, but generating dense ground truth for 3D structure requires a huge amount of time and resources. On the other hand, if the reliance on dense ground truth is relaxed, MVS systems will generalize more smoothly to new environments. To this end, we first introduce a novel semi-supervised multi-view stereo framework called a Sparse Ground truth-based MVS Network (SGT-MVSNet) that can reliably reconstruct the 3D structures even with a few ground truth 3D points. Our strategy is to divide the accurate and erroneous regions and individually conquer them based on our observation that a probability map can separate these regions. We propose a self-supervision loss called the 3D Point Consistency Loss to enhance the 3D reconstruction performance, which forces the 3D points back-projected from the corresponding pixels by the predicted depth values to meet at the same 3D co-ordinates. Finally, we propagate these improved depth pre-dictions toward edges and occlusions by the Coarse-to-fine Reliable Depth Propagation module. We generate the spare ground truth of the DTU dataset for evaluation and extensive experiments verify that our SGT-MVSNet outperforms the state-of-the-art MVS methods on the sparse ground truth setting. Moreover, our method shows comparable reconstruction results to the supervised MVS methods though we only used tens and hundreds of ground truth 3D points.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"7 1","pages":"6158-6166"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88830575","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 : 2021-10-01DOI: 10.1109/ICCV48922.2021.00304
Jung Uk Kim, Sungjune Park, Yong Man Ro
Although the visual appearances of small-scale objects are not well observed, humans can recognize them by associating the visual cues of small objects from their memorized appearance. It is called cued recall. In this paper, motivated by the memory process of humans, we introduce a novel pedestrian detection framework that imitates cued recall in detecting small-scale pedestrians. We propose a large-scale embedding learning with the large-scale pedestrian recalling memory (LPR Memory). The purpose of the proposed large-scale embedding learning is to memorize and recall the large-scale pedestrian appearance via the LPR Memory. To this end, we employ the large-scale pedestrian exemplar set, so that, the LPR Memory can recall the information of the large-scale pedestrians from the small-scale pedestrians. Comprehensive quantitative and qualitative experimental results validate the effectiveness of the proposed framework with the LPR Memory.
{"title":"Robust Small-scale Pedestrian Detection with Cued Recall via Memory Learning","authors":"Jung Uk Kim, Sungjune Park, Yong Man Ro","doi":"10.1109/ICCV48922.2021.00304","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.00304","url":null,"abstract":"Although the visual appearances of small-scale objects are not well observed, humans can recognize them by associating the visual cues of small objects from their memorized appearance. It is called cued recall. In this paper, motivated by the memory process of humans, we introduce a novel pedestrian detection framework that imitates cued recall in detecting small-scale pedestrians. We propose a large-scale embedding learning with the large-scale pedestrian recalling memory (LPR Memory). The purpose of the proposed large-scale embedding learning is to memorize and recall the large-scale pedestrian appearance via the LPR Memory. To this end, we employ the large-scale pedestrian exemplar set, so that, the LPR Memory can recall the information of the large-scale pedestrians from the small-scale pedestrians. Comprehensive quantitative and qualitative experimental results validate the effectiveness of the proposed framework with the LPR Memory.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"93 1","pages":"3030-3039"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80803930","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}
3D pose estimation has attracted increasing attention with the availability of high-quality benchmark datasets. However, prior works show that deep learning models tend to learn spurious correlations, which fail to generalize beyond the specific dataset they are trained on. In this work, we take a step towards training robust models for cross-domain pose estimation task, which brings together ideas from causal representation learning and generative adversarial networks. Specifically, this paper introduces a novel framework for causal representation learning which explicitly exploits the causal structure of the task. We consider changing domain as interventions on images under the data-generation process and steer the generative model to produce counterfactual features. This help the model learn transferable and causal relations across different domains. Our framework is able to learn with various types of unlabeled datasets. We demonstrate the efficacy of our proposed method on both human and hand pose estimation task. The experiment results show the proposed approach achieves state-of-the-art performance on most datasets for both domain adaptation and domain generalization settings.
{"title":"Learning Causal Representation for Training Cross-Domain Pose Estimator via Generative Interventions","authors":"Xiheng Zhang, Yongkang Wong, Xiaofei Wu, Juwei Lu, Mohan S. Kankanhalli, Xiangdong Li, Wei-dong Geng","doi":"10.1109/ICCV48922.2021.01108","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.01108","url":null,"abstract":"3D pose estimation has attracted increasing attention with the availability of high-quality benchmark datasets. However, prior works show that deep learning models tend to learn spurious correlations, which fail to generalize beyond the specific dataset they are trained on. In this work, we take a step towards training robust models for cross-domain pose estimation task, which brings together ideas from causal representation learning and generative adversarial networks. Specifically, this paper introduces a novel framework for causal representation learning which explicitly exploits the causal structure of the task. We consider changing domain as interventions on images under the data-generation process and steer the generative model to produce counterfactual features. This help the model learn transferable and causal relations across different domains. Our framework is able to learn with various types of unlabeled datasets. We demonstrate the efficacy of our proposed method on both human and hand pose estimation task. The experiment results show the proposed approach achieves state-of-the-art performance on most datasets for both domain adaptation and domain generalization settings.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"4 1","pages":"11250-11260"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81209676","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 : 2021-10-01DOI: 10.1109/ICCV48922.2021.00854
A. Cheraghian, Shafin Rahman, Sameera Ramasinghe, Pengfei Fang, Christian Simon, L. Petersson, Mehrtash Harandi
Few-shot class incremental learning (FSCIL) aims to incrementally add sets of novel classes to a well-trained base model in multiple training sessions with the restriction that only a few novel instances are available per class. While learning novel classes, FSCIL methods gradually forget base (old) class training and overfit to a few novel class samples. Existing approaches have addressed this problem by computing the class prototypes from the visual or semantic word vector domain. In this paper, we propose addressing this problem using a mixture of subspaces. Subspaces define the cluster structure of the visual domain and help to describe the visual and semantic domain considering the overall distribution of the data. Additionally, we propose to employ a variational autoencoder (VAE) to generate synthesized visual samples for augmenting pseudo-feature while learning novel classes incrementally. The combined effect of the mixture of subspaces and synthesized features reduces the forgetting and overfitting problem of FSCIL. Extensive experiments on three image classification datasets show that our proposed method achieves competitive results compared to state-of-the-art methods.
few -shot class incremental learning (FSCIL)的目的是在多个训练课程中，在每个类只有几个新实例可用的限制下，逐步将新类集添加到训练良好的基础模型中。在学习新类的过程中，FSCIL方法逐渐忘记了基(旧)类训练，并对少数新类样本进行过拟合。现有的方法通过从视觉或语义词向量域计算类原型来解决这个问题。在本文中，我们建议使用混合子空间来解决这个问题。子空间定义了视觉域的聚类结构，并根据数据的整体分布来描述视觉域和语义域。此外，我们建议使用变分自编码器(VAE)在增量学习新类的同时生成用于增强伪特征的合成视觉样本。混合子空间和综合特征的联合作用减少了FSCIL的遗忘和过拟合问题。在三个图像分类数据集上的大量实验表明，与现有的方法相比，我们提出的方法取得了具有竞争力的结果。
{"title":"Synthesized Feature based Few-Shot Class-Incremental Learning on a Mixture of Subspaces","authors":"A. Cheraghian, Shafin Rahman, Sameera Ramasinghe, Pengfei Fang, Christian Simon, L. Petersson, Mehrtash Harandi","doi":"10.1109/ICCV48922.2021.00854","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.00854","url":null,"abstract":"Few-shot class incremental learning (FSCIL) aims to incrementally add sets of novel classes to a well-trained base model in multiple training sessions with the restriction that only a few novel instances are available per class. While learning novel classes, FSCIL methods gradually forget base (old) class training and overfit to a few novel class samples. Existing approaches have addressed this problem by computing the class prototypes from the visual or semantic word vector domain. In this paper, we propose addressing this problem using a mixture of subspaces. Subspaces define the cluster structure of the visual domain and help to describe the visual and semantic domain considering the overall distribution of the data. Additionally, we propose to employ a variational autoencoder (VAE) to generate synthesized visual samples for augmenting pseudo-feature while learning novel classes incrementally. The combined effect of the mixture of subspaces and synthesized features reduces the forgetting and overfitting problem of FSCIL. Extensive experiments on three image classification datasets show that our proposed method achieves competitive results compared to state-of-the-art methods.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"3 1","pages":"8641-8650"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89276871","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 : 2021-10-01DOI: 10.1109/ICCV48922.2021.01374
Menghan Xia, Wenbo Hu, Xueting Liu, T. Wong
Existing halftoning algorithms usually drop colors and fine details when dithering color images with binary dot patterns, which makes it extremely difficult to recover the original information. To dispense the recovery trouble in future, we propose a novel halftoning technique that converts a color image into binary halftone with full restorability to the original version. The key idea is to implicitly embed those previously dropped information into the halftone patterns. So, the halftone pattern not only serves to reproduce the image tone, maintain the blue-noise randomness, but also represents the color information and fine details. To this end, we exploit two collaborative convolutional neural networks (CNNs) to learn the dithering scheme, under a nontrivial self-supervision formulation. To tackle the flatness degradation issue of CNNs, we propose a novel noise incentive block (NIB) that can serve as a generic CNN plug-in for performance promotion. At last, we tailor a guiding-aware training scheme that secures the convergence direction as regulated. We evaluate the invertible halftones in multiple aspects, which evidences the effectiveness of our method.
{"title":"Deep Halftoning with Reversible Binary Pattern","authors":"Menghan Xia, Wenbo Hu, Xueting Liu, T. Wong","doi":"10.1109/ICCV48922.2021.01374","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.01374","url":null,"abstract":"Existing halftoning algorithms usually drop colors and fine details when dithering color images with binary dot patterns, which makes it extremely difficult to recover the original information. To dispense the recovery trouble in future, we propose a novel halftoning technique that converts a color image into binary halftone with full restorability to the original version. The key idea is to implicitly embed those previously dropped information into the halftone patterns. So, the halftone pattern not only serves to reproduce the image tone, maintain the blue-noise randomness, but also represents the color information and fine details. To this end, we exploit two collaborative convolutional neural networks (CNNs) to learn the dithering scheme, under a nontrivial self-supervision formulation. To tackle the flatness degradation issue of CNNs, we propose a novel noise incentive block (NIB) that can serve as a generic CNN plug-in for performance promotion. At last, we tailor a guiding-aware training scheme that secures the convergence direction as regulated. We evaluate the invertible halftones in multiple aspects, which evidences the effectiveness of our method.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"82 1","pages":"13980-13989"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86647867","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 : 2021-10-01DOI: 10.1109/ICCV48922.2021.01063
Feihu Zhang, Oliver J. Woodford, V. Prisacariu, Philip H. S. Torr
Full-motion cost volumes play a central role in current state-of-the-art optical flow methods. However, constructed using simple feature correlations, they lack the ability to encapsulate prior, or even non-local knowledge. This creates artifacts in poorly constrained ambiguous regions, such as occluded and textureless areas. We propose a separable cost volume module, a drop-in replacement to correlation cost volumes, that uses non-local aggregation layers to exploit global context cues and prior knowledge, in order to disambiguate motions in these regions. Our method leads both the now standard Sintel and KITTI optical flow benchmarks in terms of accuracy, and is also shown to generalize better from synthetic to real data.
{"title":"Separable Flow: Learning Motion Cost Volumes for Optical Flow Estimation","authors":"Feihu Zhang, Oliver J. Woodford, V. Prisacariu, Philip H. S. Torr","doi":"10.1109/ICCV48922.2021.01063","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.01063","url":null,"abstract":"Full-motion cost volumes play a central role in current state-of-the-art optical flow methods. However, constructed using simple feature correlations, they lack the ability to encapsulate prior, or even non-local knowledge. This creates artifacts in poorly constrained ambiguous regions, such as occluded and textureless areas. We propose a separable cost volume module, a drop-in replacement to correlation cost volumes, that uses non-local aggregation layers to exploit global context cues and prior knowledge, in order to disambiguate motions in these regions. Our method leads both the now standard Sintel and KITTI optical flow benchmarks in terms of accuracy, and is also shown to generalize better from synthetic to real data.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"33 1","pages":"10787-10797"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91283129","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 : 2021-10-01DOI: 10.1109/ICCV48922.2021.00818
Guolei Sun, Thomas Probst, D. Paudel, Nikola Popovic, Menelaos Kanakis, Jagruti R. Patel, Dengxin Dai, L. Gool
We introduce Task Switching Networks (TSNs), a task-conditioned architecture with a single unified encoder/decoder for efficient multi-task learning. Multiple tasks are performed by switching between them, performing one task at a time. TSNs have a constant number of parameters irrespective of the number of tasks. This scalable yet conceptually simple approach circumvents the overhead and intricacy of task-specific network components in existing works. In fact, we demonstrate for the first time that multi-tasking can be performed with a single task-conditioned decoder. We achieve this by learning task-specific conditioning parameters through a jointly trained task embedding network, encouraging constructive interaction between tasks. Experiments validate the effectiveness of our approach, achieving state-of-the-art results on two challenging multi-task benchmarks, PASCAL-Context and NYUD. Our analysis of the learned task embeddings further indicates a connection to task relationships studied in the recent literature.
{"title":"Task Switching Network for Multi-task Learning","authors":"Guolei Sun, Thomas Probst, D. Paudel, Nikola Popovic, Menelaos Kanakis, Jagruti R. Patel, Dengxin Dai, L. Gool","doi":"10.1109/ICCV48922.2021.00818","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.00818","url":null,"abstract":"We introduce Task Switching Networks (TSNs), a task-conditioned architecture with a single unified encoder/decoder for efficient multi-task learning. Multiple tasks are performed by switching between them, performing one task at a time. TSNs have a constant number of parameters irrespective of the number of tasks. This scalable yet conceptually simple approach circumvents the overhead and intricacy of task-specific network components in existing works. In fact, we demonstrate for the first time that multi-tasking can be performed with a single task-conditioned decoder. We achieve this by learning task-specific conditioning parameters through a jointly trained task embedding network, encouraging constructive interaction between tasks. Experiments validate the effectiveness of our approach, achieving state-of-the-art results on two challenging multi-task benchmarks, PASCAL-Context and NYUD. Our analysis of the learned task embeddings further indicates a connection to task relationships studied in the recent literature.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"246 1","pages":"8271-8280"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76971889","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 : 2021-10-01DOI: 10.1109/ICCV48922.2021.00521
Tiantian Han, Dong Li, Ji Liu, Lu Tian, Yi Shan
Model quantization is an important mechanism for energy-efficient deployment of deep neural networks on resource-constrained devices by reducing the bit precision of weights and activations. However, it remains challenging to maintain high accuracy as bit precision decreases, especially for low-precision networks (e.g., 2-bit MobileNetV2). Existing methods have been explored to address this problem by minimizing the quantization error or mimicking the data distribution of full-precision networks. In this work, we propose a novel weight regularization algorithm for improving low-precision network quantization. Instead of constraining the overall data distribution, we separably optimize all elements in each quantization bin to be as close to the target quantized value as possible. Such bin regularization (BR) mechanism encourages the weight distribution of each quantization bin to be sharp and approximate to a Dirac delta distribution ideally. Experiments demonstrate that our method achieves consistent improvements over the state-of-the-art quantization-aware training methods for different low-precision networks. Particularly, our bin regularization improves LSQ for 2-bit MobileNetV2 and MobileNetV3-Small by 3.9% and 4.9% top-1 accuracy on ImageNet, respectively.
{"title":"Improving Low-Precision Network Quantization via Bin Regularization","authors":"Tiantian Han, Dong Li, Ji Liu, Lu Tian, Yi Shan","doi":"10.1109/ICCV48922.2021.00521","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.00521","url":null,"abstract":"Model quantization is an important mechanism for energy-efficient deployment of deep neural networks on resource-constrained devices by reducing the bit precision of weights and activations. However, it remains challenging to maintain high accuracy as bit precision decreases, especially for low-precision networks (e.g., 2-bit MobileNetV2). Existing methods have been explored to address this problem by minimizing the quantization error or mimicking the data distribution of full-precision networks. In this work, we propose a novel weight regularization algorithm for improving low-precision network quantization. Instead of constraining the overall data distribution, we separably optimize all elements in each quantization bin to be as close to the target quantized value as possible. Such bin regularization (BR) mechanism encourages the weight distribution of each quantization bin to be sharp and approximate to a Dirac delta distribution ideally. Experiments demonstrate that our method achieves consistent improvements over the state-of-the-art quantization-aware training methods for different low-precision networks. Particularly, our bin regularization improves LSQ for 2-bit MobileNetV2 and MobileNetV3-Small by 3.9% and 4.9% top-1 accuracy on ImageNet, respectively.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"102 1","pages":"5241-5250"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78164413","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 : 2021-10-01DOI: 10.1109/ICCV48922.2021.00101
M. Guo, R. Hwa, Adriana Kovashka
We propose a new approach to detect atypicality in persuasive imagery. Unlike atypicality which has been studied in prior work, persuasive atypicality has a particular purpose to convey meaning, and relies on understanding the common-sense spatial relations of objects. We propose a self-supervised attention-based technique which captures contextual compatibility, and models spatial relations in a precise manner. We further experiment with capturing common sense through the semantics of co-occurring object classes. We verify our approach on a dataset of atypicality in visual advertisements, as well as a second dataset capturing atypicality that has no persuasive intent.
{"title":"Detecting Persuasive Atypicality by Modeling Contextual Compatibility","authors":"M. Guo, R. Hwa, Adriana Kovashka","doi":"10.1109/ICCV48922.2021.00101","DOIUrl":"https://doi.org/10.1109/ICCV48922.2021.00101","url":null,"abstract":"We propose a new approach to detect atypicality in persuasive imagery. Unlike atypicality which has been studied in prior work, persuasive atypicality has a particular purpose to convey meaning, and relies on understanding the common-sense spatial relations of objects. We propose a self-supervised attention-based technique which captures contextual compatibility, and models spatial relations in a precise manner. We further experiment with capturing common sense through the semantics of co-occurring object classes. We verify our approach on a dataset of atypicality in visual advertisements, as well as a second dataset capturing atypicality that has no persuasive intent.","PeriodicalId":6820,"journal":{"name":"2021 IEEE/CVF International Conference on Computer Vision (ICCV)","volume":"5 1","pages":"952-962"},"PeriodicalIF":0.0,"publicationDate":"2021-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78278886","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：