Research map / 3D vision / Human pose estimation
Chained Predictions Using Convolutional Neural Networks
Georgia Gkioxari, Alexander Toshev, Navdeep Jaitly
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add chained predictions (condition on previous joints) vs same-depth feedforward (ablation) | 24-layer CNN · MPII val | PCKh mean (%): 75.6 -> 79.6 | Improved |
| add chained predictions to base network (ablation) | Inception-v3 encoder · MPII val | PCKh mean (%): 81.3 -> 85.3 | Improved |
| use multi-scale deception instead of single deconvolution (ablation) | 6-layer base network · MPII val | PCKh mean (%): 72.6 -> 73.9 | Improved |
| add multi-crop evaluation | Chain Model · MPII val | PCKh mean (%): 79.6 -> 82.2 | Improved |
| feed ground-truth previous joints (oracle) (ablation) | Chain Model · MPII val | PCKh mean (%): 79.6 -> 85.9 | Improved |
| use inception-v3 encoder instead of shallow 6-layer encoder | Base Network · MPII val | PCKh mean (%): 73.9 -> 81.3 | Improved |
| add chained predictions to base network (ablation) | 6-layer CNN · MPII val, wrist | PCKh (%): 69.0 -> 74.6 | Improved |
| add chaining in time (T_H=3) vs per-frame base network (ablation) | Base Network · Penn Action test | PCK mean (%): 87.5 -> 91.7 | Improved |
| connect previous outputs to hidden state (chain model vs plain conv RNN) (ablation) | convolutional RNN · Penn Action test | PCK mean (%): 90.1 -> 91.7 | Improved |
| increase time horizon from 1 to 3 (ablation) | Chain Model · Penn Action test | PCK mean (%): 91.0 -> 91.7 | Improved |
| add temporal smoothing with Viterbi | Base Network · Penn Action test | PCK mean (%): 87.5 -> 88.6 | Improved |
| add chaining in joints in addition to time (ablation) | Chain Model T_H=3 · Penn Action test | PCK mean (%): 91.7 -> 91.8 | No clear effect |
Closest papers by meaning
- Convolutional Pose Machines (CVPR 2016 · similarity 0.89)
- Convolutional Sequence to Sequence Model for Human Dynamics (CVPR 2018 · similarity 0.88)
- Deep Supervision with Intermediate Concepts (TPAMI 2018 · similarity 0.87)
- LSTM Pose Machines (CVPR 2018 · similarity 0.87)
- Adversarial Learning of Structure-Aware Fully Convolutional Networks for Landmark Localization (TPAMI 2019 · similarity 0.87)
- Structured Convolutions for Efficient Neural Network Design (NeurIPS 2020 · similarity 0.86)
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.
