Paper : https://arxiv.org/abs/2207.02425 Self-Constrained Inference Optimization on Structural Groups for Human Pose Estimation We observe that human poses exhibit strong group-wise structural correlation and spatial coupling between keypoints due to the biological constraints of different body parts. This group-wise structural correlation can be explored to improve the accuracy an arxiv.org 이번에..

Paper : https://arxiv.org/abs/2111.08557 Rethinking Keypoint Representations: Modeling Keypoints and Poses as Objects for Multi-Person Human Pose Estimation In keypoint estimation tasks such as human pose estimation, heatmap-based regression is the dominant approach despite possessing notable drawbacks: heatmaps intrinsically suffer from quantization error and require excessive computation to ge..

Paper : https://arxiv.org/pdf/2207.10387.pdf GitHub : https://github.com/luminxu/Pose-for-Everything GitHub - luminxu/Pose-for-Everything: The official repo for ECCV'22 paper: Pose for Everything: Towards Category-Agnostic Pose E The official repo for ECCV'22 paper: Pose for Everything: Towards Category-Agnostic Pose Estimation - GitHub - luminxu/Pose-for-Everything: The official repo for ECCV'2..

Paper : https://arxiv.org/pdf/2107.10466.pdf GitHub : https://github.com/IIGROUP/PoseDet GitHub - IIGROUP/PoseDet: [FG 2021] Code for PoseDet: Fast Multi-Person Pose Estimation Using Pose Embedding [FG 2021] Code for PoseDet: Fast Multi-Person Pose Estimation Using Pose Embedding - GitHub - IIGROUP/PoseDet: [FG 2021] Code for PoseDet: Fast Multi-Person Pose Estimation Using Pose Embedding github..

기존 AdaptivePose 관련 글 https://eehoeskrap.tistory.com/664 [Paper Review] AdaptivePose : Human Parts as Adaptive Points Paper : https://arxiv.org/pdf/2112.13635.pdf 이번에 리뷰할 논문은 AAAI 2022에서 소개된 "AdaptivePose : Human Parts as Adaptive Points" 라는 논문입니다. multi-person pose estimation 방법은 보통 bottom up 이나 top down 방식으 eehoeskrap.tistory.com Paper : https://arxiv.org/pdf/2210.04014v1.pdf AdaptivePose에 이어 A..

Paper : https://arxiv.org/pdf/2112.13635.pdf Github : https://github.com/buptxyb666/AdaptivePose GitHub - buptxyb666/AdaptivePose: This is an official implementation of our AAAI2022 paper“AdaptivePose: Human Parts as Adapti This is an official implementation of our AAAI2022 paper“AdaptivePose: Human Parts as Adaptive Points” - GitHub - buptxyb666/AdaptivePose: This is an official implementation ..

Paper : https://arxiv.org/abs/2208.00090 Explicit Occlusion Reasoning for Multi-person 3D Human Pose Estimation Occlusion poses a great threat to monocular multi-person 3D human pose estimation due to large variability in terms of the shape, appearance, and position of occluders. While existing methods try to handle occlusion with pose priors/constraints, data augme arxiv.org GitHub : https://gi..

Paper : https://openaccess.thecvf.com/content/CVPR2022/papers/Kundu_Uncertainty-Aware_Adaptation_for_Self-Supervised_3D_Human_Pose_Estimation_CVPR_2022_paper.pdf 본 논문의 main contribution은 아래와 같습니다. multi-representation pose network를 사용하는 MRP-Net을 제안하였으며, pose-uncertainty는 두 가지 다양한 설계(model-free, model-based)를 기반으로 하는 2개의 output head를 통한 pose prediction간의 불일치로 정량화됩니다. 제안된 포즈와 joint uncertainty의 효율..

Paper : https://openaccess.thecvf.com/content_ICCV_2019/papers/Cheng_Occlusion-Aware_Networks_for_3D_Human_Pose_Estimation_in_Video_ICCV_2019_paper.pdf 본 논문에서는 3D Human Pose Estimation에서 Occlusion 문제를 해결하기 위하여 occlusion aware deep learning framework 제안합니다. 이를 위해 keypoint의 2D confidence heatmap과 optical flow의 consistency constraint를 사용하여 occluded keypoint의 unreliable estimation을 filtering 합니다. oc..

Paper : https://arxiv.org/abs/2107.03215 FasterPose: A Faster Simple Baseline for Human Pose Estimation The performance of human pose estimation depends on the spatial accuracy of keypoint localization. Most existing methods pursue the spatial accuracy through learning the high-resolution (HR) representation from input images. By the experimental analysis, w arxiv.org 2021년 7월에 아카이브에서 공개된 Faster..

오늘은 Pose Estimation 분야에서 CPN으로 유명한 Casecaded Pyramid Network 논문을 빠르게 훑어보도록 하겠습니다. 이 논문에서는 여러가지 복합적인 요인(Occlusion, Background, ...)에 의하여 검출하기 어려운 키포인트들을 잘 검출하기 위해 CPN이라는 구조를 제안했습니다. 크게 아래 그림과 같이 GlobalNet과 RefineNet으로 구성되어있습니다. 이름에서도 알 수 있듯이 GlobalNet에서는 feature pyramid network 기반으로 global feature를 추출하여 localization하는 방식이며, RefineNet은 GlobalNet에서 localization 하지 못한 hard keypoint를 hard keypoint mi..

이번에 읽을 논문은 ECCV 2022에서 발표 예정인 SmoothNet 이라는 논문 입니다. 제가 개인적으로 Plug-and-Play Network를 굉장히 좋아합니다 ㅎㅎㅎㅎ 제목부터 마음에 드네요. Abstract 내용을 간단히 소개하자면 human motion video 환경에서 기존 pose estimation 방식들의 output jitter들은 프레임 전반에 걸쳐 다양한 estimation error를 발생시키는데, 거의 보이지 않거나 가려진 동작의 경우 여러 관절의 추정 위치가 연속 프레임 시퀀스에 대한 실제 값에서 크게 벗어나 significant jitter들을 발생시키게 됩니다. 이러한 문제를 해결하기 위해 SmoothNet을 제안했습니다. 본 논문에서는 jitter를 완화하기 위해 기..