Research map / Detection and segmentation / CNN features
DeepVoting: A Robust and Explainable Deep Network for Semantic Part Detection Under Partial Occlusion
Read the paperCheck 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 |
|---|---|---|---|
| Add context outside object bounding boxes during training (ablation) | DeepVoting · VehicleSemanticPart, no occlusions | mAP (%): 72.0 -> 74.0 | Helped |
| Add context outside object bounding boxes during training (ablation) | DeepVoting · VehicleOcclusion L1 | mAP (%): 53.7 -> 58.0 | Helped |
| Add context outside object bounding boxes during training (ablation) | DeepVoting · VehicleOcclusion L2 | mAP (%): 42.6 -> 46.9 | Helped |
| Add context outside object bounding boxes during training (ablation) | DeepVoting · VehicleOcclusion L3 | mAP (%): 31.6 -> 35.2 | Helped |
| Replace standalone voting with end-to-end joint training of concept and voting layers | DeepVoting · VehicleSemanticPart, no occlusions | mAP (%): 53.1 -> 72.0 | Helped |
| Use voting over multiple visual cues rather than one visual concept directly | Voting · VehicleSemanticPart, no occlusions | mAP (%): 30.4 -> 53.1 | Helped |
| Provide ground-truth scale instead of predicted scale (ablation) | DeepVoting · VehicleSemanticPart, no occlusions | mAP (%): 72.0% -> 74.5% | Helped |
| Shrink voting kernel from 15×15 to 11×11 (ablation) | DeepVoting · VehicleOcclusion L3 | mAP (%): 31.6% -> 30.6% | Hurt |
| Increase voting kernel from 15×15 to 19×19 (ablation) | DeepVoting · VehicleOcclusion L3 | mAP (%): 31.6% -> 31.8% | Helped |
| Add ground-truth bounding boxes as Faster-RCNN proposals (ablation) | Faster-RCNN · VehicleOcclusion L3 | mAP (%): 23.0 -> 24.2 | Helped |
| Train and test on occlusion-free images, then transfer learned visual-cue relationships to occluded images | DeepVoting · VehicleOcclusion L3 | mAP (%): 31.6 vs. 23.0 | Helped |
Closest papers by meaning
- Robust Object Detection Under Occlusion With Context-Aware CompositionalNets (CVPR 2020 · similarity 0.88)
- Growing Interpretable Part Graphs on ConvNets via Multi-Shot Learning (AAAI 2017 · similarity 0.87)
- Deep Supervision With Shape Concepts for Occlusion-Aware 3D Object Parsing (CVPR 2017 · similarity 0.87)
- Objects as Context for Detecting Their Semantic Parts (CVPR 2018 · similarity 0.87)
- Beyond Holistic Object Recognition: Enriching Image Understanding With Part States (CVPR 2018 · similarity 0.87)
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · similarity 0.85)
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.