Online Knowledge Distillation for Efficient Pose Estimation

Li, Zheng; Ye, Jingwen; Song, Mingli; Huang, Ying; Pan, Zhigeng

doi:10.1109/iccv48922.2021.01153

Cited by 77 publications

(36 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We think the generalization ability of the model is more important for the dehazing tasks, and a little distortion is tolerable. ); Variant B, the proposed method without the FAB (we replace it to feature aggregation unit designed in paper 30 ); Variant C, the proposed method without the multiscale feature shared network (we replace it to a single scale network consisted of AGRDBs ); Variant D, the proposed method without the knowledge distillation of the model-based student branch; Variation E, the proposed method without the knowledge distillation of the model-free student branch; Variation F, the proposed method. We train these variants on the ITS datasets for 30 epochs and test the trained variants on the SOTS outdoor datasets, I-HAZE, and O-HAZE, and conduct quantitative comparisons to evaluate the performance of each variant.…”

Section: Comparisons With State-of-the-art Methodsmentioning

confidence: 99%

“…Zhang et al 29 proposed a self-distillation strategy, which constructs a deep CNN and distills the features of deep convolutions to the shallower convolutions. Moreover, Li et al 30 proposed a novel online knowledge distillation method, which does not rely on pretrained teachers and improves the accuracy of pose estimation. Inspired by it, we build an online knowledge distillation network for single image dehazing named OKDNet.…”

Section: Knowledge Distillationmentioning

confidence: 99%

See 1 more Smart Citation

Online knowledge distillation network for single image dehazing

Lan

Cui

et al. 2022

Sci Rep

View full text Add to dashboard Cite

Single image dehazing, as a key prerequisite of high-level computer vision tasks, catches more and more attentions. Traditional model-based methods recover haze-free images via atmospheric scattering model, which achieve favorable dehazing effect but endure artifacts, halos, and color distortion. By contrast, recent learning-based methods dehaze images by a model-free way, which achieve better color fidelity but tend to acquire under-dehazed results due to lacking of knowledge guiding. To combine these merits, we propose a novel online knowledge distillation network for single image dehazing named OKDNet. Specifically, the proposed OKDNet firstly preprocesses hazy images and acquires abundant shared features by a multiscale network constructed with attention guided residual dense blocks. After that, these features are sent to different branches to generate two preliminary dehazed images via supervision training: one branch acquires dehazed images via the atmospheric scattering model; another branch directly establishes the mapping relationship between hazy images and clear images, which dehazes images by a model-free way. To effectively fuse useful information from these two branches and acquire a better dehazed results, we propose an efficient feature aggregation block consisted of multiple parallel convolutions with different receptive. Moreover, we adopt a one-stage knowledge distillation strategy named online knowledge distillation to joint optimization of our OKDNet. The proposed OKDNet achieves superior performance compared with state-of-the-art methods on both synthetic and real-world images with fewer model parameters. Project website: https://github.com/lanyunwei/OKDNet.

show abstract

Section: Comparisons With State-of-the-art Methodsmentioning

confidence: 99%

Section: Knowledge Distillationmentioning

confidence: 99%

Online knowledge distillation network for single image dehazing

Lan

Cui

et al. 2022

Sci Rep

View full text Add to dashboard Cite

show abstract

“…Designing efficient human pose estimators has been intensively studied for practical usage. For extracting 2D pose from images, state-of-the-art methods [22,26,36,50,53] have achieved real-time inference speed. In terms of multiview 3D pose estimation, Bultman et al .…”

Section: Efficient Human Pose Estimationmentioning

confidence: 99%

Faster VoxelPose: Real-time 3D Human Pose Estimation by Orthographic Projection

Ye¹,

Zhu²,

Wang³

et al. 2022

Preprint

View full text Add to dashboard Cite

While the voxel-based methods have achieved promising results for multi-person 3D pose estimation from multi-cameras, they suffer from heavy computation burdens, especially for large scenes. We present Faster VoxelPose to address the challenge by re-projecting the feature volume to the three two-dimensional coordinate planes and estimating X, Y, Z coordinates from them separately. To that end, we first localize each person by a 3D bounding box by estimating a 2D box and its height based on the volume features projected to the xy-plane and z-axis, respectively. Then for each person, we estimate partial joint coordinates from the three coordinate planes separately which are then fused to obtain the final 3D pose. The method is free from costly 3D-CNNs and improves the speed of VoxelPose by ten times and meanwhile achieves competitive accuracy as the state-of-the-art methods, proving its potential in real-time applications.

show abstract

“…More recently, the deep learning techniques [81,90,103,146,147] enable learning feature representations automatically from data, which has significantly contributed to the advancement of human pose estimation. These deep learningbased approaches [9,86,91,99,101,153,161,181], commonly building upon the success of convolutional neural networks, have achieved outstanding performance on this task. Given the rapid development, this paper seeks to track recent progress and summarize their accomplishments to deliver a clearer panorama for 2D human pose estimation.…”

Section: Introductionmentioning

confidence: 99%

2D Human Pose Estimation: A Survey

Chen¹,

Feng²,

Wu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Human pose estimation aims at localizing human anatomical keypoints or body parts in the input data (e.g., images, videos, or signals). It forms a crucial component in enabling machines to have an insightful understanding of the behaviors of humans, and has become a salient problem in computer vision and related fields. Deep learning techniques allow learning feature representations directly from the data, significantly pushing the performance boundary of human pose estimation. In this paper, we reap the recent achievements of 2D human pose estimation methods and present a comprehensive survey. Briefly, existing approaches put their efforts in three directions, namely network architecture design, network training refinement, and post processing. Network architecture design looks at the architecture of human pose estimation models, extracting more robust features for keypoint recognition and localization. Network training refinement tap into the training of neural networks and aims to improve the representational ability of models. Post processing further incorporates model-agnostic polishing strategies to improve the performance of keypoint detection. More than 200 research contributions are involved in this survey, covering methodological frameworks, common benchmark datasets, evaluation metrics, and performance comparisons. We seek to provide researchers with a more comprehensive and systematic review on human pose estimation, allowing them to acquire a grand panorama and better identify future directions. ‡ The first two authors have equal contribution.

show abstract

Online Knowledge Distillation for Efficient Pose Estimation

Cited by 77 publications

References 43 publications

Online knowledge distillation network for single image dehazing

Online knowledge distillation network for single image dehazing

Faster VoxelPose: Real-time 3D Human Pose Estimation by Orthographic Projection

2D Human Pose Estimation: A Survey

Contact Info

Product

Resources

About