Research map / Detection and segmentation / Open-set and few-shot detection
A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection
Xiaolong Wang, Abhinav Shrivastava, Abhinav Gupta
Check 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 adversarial feature occlusions and deformations | Fast R-CNN with VGG · VOC 2007 test | mAP: 69.1 -> 71.4 | Improved |
| Add adversarial feature occlusions and deformations | Fast R-CNN with VGG · VOC 2012 test | mAP: 66.4 -> 69.0 | Improved |
| Add adversarial feature occlusions and deformations | Fast R-CNN with VGG · MS COCO minival | AP50 (VOC metric): 42.7 -> 46.2 | Improved |
| Add adversarial feature occlusions and deformations | Fast R-CNN with VGG · MS COCO minival | AP: 25.7 -> 27.1 | Improved |
| Add adversarial feature occlusions and deformations | Fast R-CNN with ResNet · VOC 2007 test | mAP: 71.8 -> 73.6 | Improved |
| Combine ASDN occlusion with ASTN deformation (ablation) | Fast R-CNN with AlexNet · VOC 2007 test | mAP: 58.5 -> 58.9 | Improved |
| Replace adversarial feature occlusions with random spatial dropout (ablation) | ASDN-based Fast R-CNN with AlexNet · VOC 2007 test | mAP: 58.5 -> 57.3 | Hurt |
| Replace ASDN with exhaustive hard occlusion-window search (ablation) | ASDN-based Fast R-CNN with AlexNet · VOC 2007 test | mAP: 58.5 -> 57.7 | Hurt |
| Fix ASDN parameters instead of jointly learning them (ablation) | ASDN-based Fast R-CNN with AlexNet · VOC 2007 test | mAP: 58.5 -> 57.5 | Hurt |
| Jointly learn ASDN with the detector (ablation) | Fast R-CNN with AlexNet · VOC 2007 test | mAP: 57.0 -> 58.5 | Improved |
| Use ASTN adversarial deformations instead of random proposal jittering (ablation) | Fast R-CNN with AlexNet · VOC 2007 test | mAP: 57.3 -> 58.1 | Improved |
| Use A-Fast-RCNN instead of OHEM | VGG detector · VOC 2007 test | mAP: 69.9 -> 71.4 | Improved |
Closest papers by meaning
- Increasing CNN Robustness to Occlusions by Reducing Filter Support (ICCV 2017 · similarity 0.87)
- Generative Adversarial Learning Towards Fast Weakly Supervised Detection (CVPR 2018 · similarity 0.87)
- Robust Object Detection Under Occlusion With Context-Aware CompositionalNets (CVPR 2020 · similarity 0.86)
- Spatially Invariant Unsupervised Object Detection with Convolutional Neural Networks (AAAI 2019 · similarity 0.86)
- Domain Adaptive Object Detection via Asymmetric Tri-way Faster-RCNN (ECCV 2020 · similarity 0.86)
- Straight to Shapes: Real-Time Detection of Encoded Shapes (CVPR 2017 · 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.
