Research map / Detection and segmentation / CNN features
Inside-Outside Net: Detecting Objects in Context With Skip Pooling and Recurrent Neural Networks
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 |
|---|---|---|---|
| Add multi-layer ROI pooling with feature normalization (ablation) | Fast R-CNN · PASCAL VOC 2007 test | mAP: 70.8 -> 74.6 | Helped |
| Normalize pooled features before combining layers (ablation) | ION multi-layer pooling · PASCAL VOC 2007 test | mAP: 63.6 -> 74.6 | Helped |
| Add semantic segmentation loss (ablation) | ION · PASCAL VOC 2007 test | mAP: 75.6 -> 76.5 | Helped |
| Use two stacked 4-directional IRNN layers for context instead of two 3x3 convolution layers (ablation) | ION · PASCAL VOC 2007 test | mAP: 74.8 -> 75.6 | Helped |
| Replace IRNN context with global average pooling (ablation) | ION · PASCAL VOC 2007 test | mAP: 75.6 -> 74.9 | Hurt |
| Replace learned recurrent transition Whh with identity (ablation) | ION IRNN · PASCAL VOC 2007 test | mAP: 76.5 -> 76.1 | No clear effect |
| Vary the feature-normalization scheme among the tested scales and reductions (ablation) | ION · PASCAL VOC 2007 test | mAP: 76.2; 76.4; 76.5; 76.6 | No clear effect |
| Add two rounds of bounding-box regression and weighted voting (ablation) | ION · PASCAL VOC 2007 test | mAP: 76.5% to 78.5% | Helped |
| Remove dropout (ablation) | ION · PASCAL VOC 2007 test | mAP: 78.5% to 79.2% | Helped |
| Add box voting (ablation) | ION · MS COCO 2015 test-dev | AP (IoU 0.5:0.95): -0.3 mAP | Hurt |
| Add context, multi-scale features, and the ION detector design | Fast R-CNN · MS COCO 2015 test-dev | AP (IoU 0.5:0.95): 20.5 -> 24.9 | Helped |
| Add the ION architecture and training improvements | MR-CNN · PASCAL VOC 2012 test | mAP: 73.9 -> 77.9 | Helped |
Closest papers by meaning
- Detect Globally, Refine Locally: A Novel Approach to Saliency Detection (CVPR 2018 · similarity 0.89)
- Spatially Invariant Unsupervised Object Detection with Convolutional Neural Networks (AAAI 2019 · similarity 0.88)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.87)
- R-FCN++: Towards Accurate Region-Based Fully Convolutional Networks for Object Detection (AAAI 2018 · similarity 0.87)
- Seeing without Looking: Contextual Rescoring of Object Detections for AP Maximization (CVPR 2020 · similarity 0.87)
- Recognize Any Regions (NeurIPS 2024 · 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.