Research map / Detection and segmentation / CNN features
Learning Rich Features at High-Speed for Single-Shot Object Detection
Tiancai Wang, Rao Muhammad Anwer, Hisham Cholakkal, Fahad Shahbaz Khan, Yanwei Pang, Ling Shao
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 light-weight scratch network (LSN) (ablation) | SSD (VGG-16) · MS COCO minival (300x300, VGG-16) | COCO AP: 25.3 -> 28.9 | Improved |
| add LSN (ablation) | SSD (VGG-16) · MS COCO minival (300x300, VGG-16) | COCO APs: 6.2 -> 10.6 | Improved |
| add bi-directional network on top of LSN (ablation) | SSD + LSN · MS COCO minival (300x300, VGG-16) | COCO AP: 28.9 -> 31.9 | Improved |
| use LSN features trained from scratch instead of reusing backbone conv1 (ablation) | SSD (VGG-16) · MS COCO minival (300x300, VGG-16) | COCO AP: 26.8 -> 28.9 | Improved |
| use our top-down scheme instead of FPN top-down (ablation) | SSD (VGG-16) · MS COCO minival (300x300, VGG-16) | COCO AP: 27.3 -> 29.3 | Improved |
| use dense top-down instead of cascade top-down with cascade bottom-up (ablation) | bi-directional network · MS COCO minival (300x300, VGG-16, bi-directional network only, no LSN) | COCO AP: 31.0 -> 31.9 | Improved |
| use cascade bottom-up instead of dense bottom-up with cascade top-down (ablation) | bi-directional network · MS COCO minival (300x300, VGG-16, bi-directional network only, no LSN) | COCO AP: 29.6 -> 31.9 | Improved |
| use dense connections in both schemes instead of cascade bottom-up and dense top-down (ablation) | bi-directional network · MS COCO minival (300x300, VGG-16, bi-directional network only, no LSN) | COCO AP: 31.9 -> 31.3 | Mixed |
| replace SSD with our detector (VGG-16, 300x300) | SSD (VGG-16) · MS COCO test-dev (300x300) | COCO AP: 25.3 -> 32.0 | Improved |
| replace SSD with our detector (VGG-16, 512x512) | SSD (VGG-16) · MS COCO test-dev (512x512) | COCO AP: 28.8 -> 36.2 | Improved |
| replace SSD with our detector on UAV data | SSD (VGG-16) · UAVDT test (512x512 or 1024x540) | AP: 33.62 -> 37.81 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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
- Faster-RCNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 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
- ScratchDet: Training Single-Shot Object Detectors From Scratch (CVPR 2019 · similarity 0.91)
- Region-Based Convolutional Networks for Accurate Object Detection and Segmentation (TPAMI 2016 · similarity 0.90)
- "A Simple Approach and Benchmark for 21,000-Category Object Detection" (ECCV 2022 · similarity 0.90)
- M2Det: A Single-Shot Object Detector Based on Multi-Level Feature Pyramid Network (AAAI 2019 · similarity 0.90)
- Towards Large-Scale Small Object Detection: Survey and Benchmarks (TPAMI 2023 · similarity 0.90)
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.90)
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.
