Research map / Architectures and efficiency / CNN design
FCPose: Fully Convolutional Multi-Person Pose Estimation With Dynamic Instance-Aware Convolutions
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use dynamic instance-aware keypoint heads and remove ROI-based processing | FCPose with ResNet-50 · COCO test-dev | AP^kp (%): 63.9 -> 64.3 | Helped |
| Use dynamic instance-aware keypoint heads and remove ROI-based processing | FCPose with ResNet-50 · COCO test-dev | inference time (ms): 89 -> 68 | Helped |
| Use a regression-based keypoint refinement module (ablation) | FCPose · COCO val2017 | AP^kp: 56.2 -> 63.0 | Helped |
| Use a regression-based keypoint refinement module instead of deconvolution upsampling (ablation) | FCPose · COCO val2017 | AP^kp: 60.1 -> 63.0 | Helped |
| Use deconvolution upsampling instead of the proposed refinement module (ablation) | FCPose · COCO val2017 | inference time (ms): 71 -> 135 | Hurt |
| Share one refinement module across instances (ablation) | FCPose · COCO val2017 | AP^kp: 62.7 -> 63.0 | Helped |
| Vary input feature channels from the 32-channel default (ablation) | FCPose dynamic keypoint head · COCO val2017 | AP^kp: 62.8, 63.0, 62.6 | Mixed |
| Vary dynamic keypoint head depth from 3 convolution layers (ablation) | FCPose dynamic keypoint head · COCO val2017 | AP^kp: 62.8, 63.0, 62.6 | Mixed |
| Vary dynamic keypoint head width from 32 channels (ablation) | FCPose dynamic keypoint head · COCO val2017 | AP^kp: 62.5, 63.0, 62.6 | Mixed |
| Use a DLA-34 backbone for the real-time FCPose configuration | FCPose · COCO test-dev | AP^kp (%): 64.2 -> 64.8 | Helped |
Closest papers by meaning
- Towards Accurate Multi-Person Pose Estimation in the Wild (CVPR 2017 · similarity 0.90)
- Simple Pose: Rethinking and Improving a Bottom-up Approach for Multi-Person Pose Estimation (AAAI 2020 · similarity 0.90)
- DeepCut: Joint Subset Partition and Labeling for Multi Person Pose Estimation (CVPR 2016 · similarity 0.89)
- QueryPose: Sparse Multi-Person Pose Regression via Spatial-Aware Part-Level Query (NeurIPS 2022 · similarity 0.88)
- End-to-End Multi-Person Pose Estimation with Pose-Aware Video Transformer (AAAI 2026 · similarity 0.88)
- AdaptivePose: Human Parts as Adaptive Points (AAAI 2022 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.