Research map / Detection and segmentation / CNN features
Receptive Field Block Net for Accurate and Fast Object Detection
Songtao Liu, Di Huang, andYunhong Wang
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 |
|---|---|---|---|
| replace last conv layer with RFB-max pooling (ablation) | SSD300* · VOC 2007 test (07+12) | mAP (%): 77.2 -> 79.1 | Improved |
| tune RFB parameters to cortex size-eccentricity ratio (RFB-s) (ablation) | RFB max pooling · VOC 2007 test (07+12) | mAP (%): 79.1 -> 79.6 | Improved |
| tune RFB parameters to cortex size-eccentricity ratio (ablation) | RFB dilated conv · VOC 2007 test (07+12) | mAP (%): 80.1 -> 80.5 | Improved |
| add 6 default priors at conv4_3 (ablation) | RFB · VOC 2007 test (07+12) | mAP (%): 79.6 -> 79.8 | Improved |
| add 6 default priors at conv4_3 (ablation) | original SSD · VOC 2007 test (07+12) | mAP (%) | No clear effect |
| use dilated convolution instead of dilated pooling (ablation) | RFB · VOC 2007 test (07+12) | mAP (%): 79.8 -> 80.5 | Improved |
| use RFB-avg pooling instead of max pooling (ablation) | RFB · VOC 2007 test (07+12) | mAP (%): 79.8 -> 79.8 | No clear effect |
| replace Inception block with RFB (ablation) | SSD · VOC 2007 test | mAP (%): 78.4 -> 80.1 | Improved |
| replace Inception block with RFB (ablation) | SSD · COCO minival | mAP (%): 27.3 -> 29.7 | Improved |
| replace ASPP-S with RFB (ablation) | SSD · COCO minival | mAP (%): 28.1 -> 29.7 | Improved |
| replace Deformable CNN block with RFB (ablation) | SSD · VOC 2007 test | mAP (%): 79.5 -> 80.1 | Improved |
| add RFB to MobileNet backbone | SSD300 MobileNet · COCO minival | mAP (%): 19.3% -> 20.7% | 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 R-CNN — 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
- RF-Net: An End-To-End Image Matching Network Based on Receptive Field (CVPR 2019 · similarity 0.87)
- RecursiveDet: End-to-End Region-Based Recursive Object Detection (ICCV 2023 · similarity 0.86)
- NAS-FPN: Learning Scalable Feature Pyramid Architecture for Object Detection (CVPR 2019 · similarity 0.86)
- Representation Sharing for Fast Object Detector Search and Beyond (ECCV 2020 · similarity 0.86)
- Receptive Fields As Experts in Convolutional Neural Architectures (ICML 2024 · similarity 0.85)
- Mamba YOLO: A Simple Baseline for Object Detection with State Space Model (AAAI 2025 · similarity 0.85)
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.
