Research map / Detection and segmentation / CNN features
CBNet: A Novel Composite Backbone Network Architecture for Object Detection
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 Dual-Backbone CBNet with AHLC | FPN + ResNet101 · COCO test-dev | AP_box: 39.4 -> 41.0 | Helped |
| Add Triple-Backbone CBNet with AHLC | FPN + ResNet101 · COCO test-dev | AP_box: 39.4 -> 41.7 | Helped |
| Add Dual-Backbone CBNet with AHLC | Mask R-CNN + ResNet101 · COCO test-dev | AP_mask: 35.9 -> 37.0 | Helped |
| Add Triple-Backbone CBNet with AHLC | Cascade R-CNN + ResNet101 · COCO test-dev | AP_box: 42.8 -> 44.9 | Helped |
| Add Triple-Backbone CBNet with AHLC | Cascade Mask R-CNN + ResNeXt152 · COCO test-dev | AP_mask: 41.0 -> 43.3 | Helped |
| Use Same Level Composition instead of AHLC (ablation) | Dual-ResNet101 CBNet with FPN · COCO test-dev | AP_box: 41.0 -> 38.9 | Hurt |
| Use Adjacent Lower-Level Composition instead of AHLC (ablation) | Dual-ResNet101 CBNet with FPN · COCO test-dev | AP_box: 41.0 -> 36.5 | Hurt |
| Use Adjacent Dense Level Composition instead of AHLC (ablation) | Dual-ResNet101 CBNet with FPN · COCO test-dev | AP_box: 41.0 -> 40.7 | Hurt |
| Add adjacent higher-level composite connections (AHLC) (ablation) | Dual-ResNet101 CBNet with FPN · COCO test-dev | AP_box: 39.4 -> 41.0 | Helped |
| Share weights between the two backbones (ablation) | Dual-ResNet101 CBNet with FPN · COCO test-dev | AP_box: 39.4 -> 40.4 | Helped |
| Use separate weights rather than sharing weights between backbones (ablation) | Dual-ResNet101 CBNet with FPN · COCO test-dev | AP_box: 40.4 -> 41.0 | Helped |
| Remove the two early stages of the Assistant Backbone (ablation) | Dual-ResNet101 CBNet with FPN · COCO test-dev | AP_box: 41.0 -> 40.8 | Hurt |
Closest papers by meaning
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.88)
- DetNAS: Backbone Search for Object Detection (NeurIPS 2019 · similarity 0.88)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.87)
- Battle of the Backbones: A Large-Scale Comparison of Pretrained Models across Computer Vision Tasks (NeurIPS 2023 · similarity 0.87)
- M2Det: A Single-Shot Object Detector Based on Multi-Level Feature Pyramid Network (AAAI 2019 · similarity 0.87)
- BFBox: Searching Face-Appropriate Backbone and Feature Pyramid Network for Face Detector (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.