Research map / Detection and segmentation / CNN features
Deep Feature Pyramid Reconfiguration for Object Detection
Tao Kong, Fuchun Sun, Chuanqi Tan, Huaping Liu, Wenbing Huang
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 |
|---|---|---|---|
| add global attention (SE block) to local reconfiguration (ablation) | SSD+Local · VOC2007 SSD300 | mAP (%): 79.0 -> 79.6 | Improved |
| add residual term x_l (not only R) in local reconfiguration (ablation) | SSD+Local · VOC2007 SSD300 | mAP (%): 78.6 -> 79.0 | Improved |
| use whole feature hierarchy instead of only deeper layers (ablation) | SSD+Global-Local · VOC2007 SSD300 | mAP (%): 79.4 -> 79.6 | Improved |
| add lateral connections to SSD | SSD · VOC2007 SSD300 | mAP (%): 77.5 -> 78.5 | Improved |
| replace lateral connections with global-local reconfiguration | SSD+lateral · VOC2007 SSD300 | mAP (%): 78.5 -> 79.6 | Improved |
| replace lateral connections with global-local reconfiguration | SSD+lateral · VOC2007 SSD300 | FPS: 37 -> 39.5 | Improved |
| add global-local reconfiguration to SSD | SSD · VOC2007 SSD300 | FPS: 44 -> 39.5 | Hurt |
| replace FPN lateral connections with global-local | Faster R-CNN ResNet-101 · VOC2007 test | mAP (%): 79.8 -> 80.6 | Improved |
| replace FPN lateral connections with global-local | Faster R-CNN ResNet-50 · VOC2007 test | mAP (%): 78.8 -> 79.4 | Improved |
| use ResNet-101 instead of VGG-16 backbone with 512 input | SSD Ours512 · VOC2007 test | mAP (%): 81.1 -> 82.4 | Improved |
| enlarge input from 300 to 512 | SSD Ours VGG-16 · MS COCO test-dev | AP 0.5:0.95 (%): 28.4 -> 31.5 | Improved |
| add global-local reconfiguration to SSD513 | SSD513 · MS COCO small objects | AP_s (%): 10.2 -> 14.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- R-FCN — R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016) · better on 1 of 1 shared result
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
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.91)
- Parallel Feature Pyramid Network for Object Detection (ECCV 2018 · similarity 0.88)
- DetectoRS: Detecting Objects With Recursive Feature Pyramid and Switchable Atrous Convolution (CVPR 2021 · similarity 0.88)
- M2Det: A Single-Shot Object Detector Based on Multi-Level Feature Pyramid Network (AAAI 2019 · similarity 0.88)
- Transformer-Based Context Condensation for Boosting Feature Pyramids in Object Detection (IJCV 2023 · similarity 0.88)
- AugFPN: Improving Multi-Scale Feature Learning for Object Detection (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.
