Research map / Detection and segmentation / CNN features
Towards Interpretable Object Detection by Unfolding Latent Structures
Tianfu Wu, Xi Song
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 |
|---|---|---|---|
| use top-k sparsity value-network policy instead of baseline weights (ablation) | AOG5,5,1 · PASCAL VOC2007 test | AP@0.5: 82.1 -> 80.4 | Hurt |
| use adversarial-attack value-network policy instead of baseline weights (ablation) | AOG5,5,1 · PASCAL VOC2007 test | AP@0.5: 82.1 -> 81.4 | Hurt |
| use top-k policy instead of baseline weights (ablation) | AOG5,5,1 · COCO val2017 | AP@0.5: 38.2 -> 37.0 | Hurt |
| use adversarial policy instead of baseline weights (ablation) | AOG5,5,1 · COCO val2017 | AP@0.5: 38.2 -> 38.0 | Hurt |
| use top-k policy instead of baseline weights (ablation) | AOG3,3,1 · PASCAL VOC2007 test | AP@0.5: 81.9 -> 81.2 | Hurt |
| use adversarial policy instead of top-k policy (ablation) | AOG5,5,1 · PASCAL VOC2007 test | AP@0.5: 80.4 -> 81.4 | Improved |
| use adversarial policy instead of baseline weights (ablation) | AOG7,7,1 · PASCAL VOC2007 test | AP@0.5: 81.7 -> 81.7 | No clear effect |
| use grid 5,5,1 AOG instead of 3,3,1 (base policy) (ablation) | AOG + base · PASCAL VOC2007 test | AP@0.5: 81.9 -> 82.1 | Improved |
| use grid 7,7,1 AOG instead of 5,5,1 (base policy) (ablation) | AOG + base · PASCAL VOC2007 test | AP@0.5: 82.1 -> 81.7 | Hurt |
| replace Faster R-CNN with AOG5,5,1 + base interpretable head | Faster R-CNN · PASCAL VOC2007 test | AP@0.5: 82.1 -> 82.1 | No clear effect |
| replace Faster R-CNN with AOG5,5,1 + base interpretable head | Faster R-CNN · COCO val2017 | AP@0.5: 38.5 -> 38.2 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Faster R-CNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 0 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Spatially Invariant Unsupervised Object Detection with Convolutional Neural Networks (AAAI 2019 · similarity 0.89)
- Network Dissection: Quantifying Interpretability of Deep Visual Representations (CVPR 2017 · similarity 0.88)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.87)
- RelationNet++: Bridging Visual Representations for Object Detection via Transformer Decoder (NeurIPS 2020 · similarity 0.87)
- Object Detection Based on Region Decomposition and Assembly (AAAI 2019 · similarity 0.87)
- A Disentangling Invertible Interpretation Network for Explaining Latent Representations (CVPR 2020 · similarity 0.87)
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.
