Research map / Detection and segmentation / CNN features
FishNet: A Versatile Backbone for Image, Region, and Pixel Level Prediction
Shuyang Sun, Jiangmiao Pang, Jianping Shi, Shuai Yi, Wanli Ouyang
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 |
|---|---|---|---|
| Use FishNet-150 instead of ResNet-50 | Image classification · ImageNet-1k val | Top-1 error: 23.78% -> 21.93% | Improved |
| Use FishNet-150 instead of ResNet-101 | Image classification · ImageNet-1k val | Top-1 error: 22.30% -> 21.93% | Improved |
| Use FishNeXt-150 (4d) instead of ResNeXt-50 (32×4d) | Image classification · ImageNet-1k val | Top-1 error: 22.2% -> 21.5% | Improved |
| Use FishNet-150 instead of ResNet-50† | Mask R-CNN instance segmentation · MS COCO val-2017 | AP: +2.3 | Improved |
| Use FishNet-150 instead of ResNeXt-50† | Mask R-CNN instance segmentation · MS COCO val-2017 | AP: +1.3 | Improved |
| Use FishNet-150 instead of ResNet-50† | Mask R-CNN object detection · MS COCO val-2017 | AP: +2.8 | Improved |
| Use FishNet-150 instead of ResNeXt-50† | Mask R-CNN object detection · MS COCO val-2017 | AP: +1.5 | Improved |
| Use FishNet-150 instead of ResNet-50† | FPN object detection · MS COCO val-2017 | AP: +2.6 | Improved |
| Use FishNet-150 instead of ResNeXt-50† | FPN object detection · MS COCO val-2017 | AP: +1.3 | Improved |
| Use 2×2 max-pooling instead of 3×3 max-pooling (ablation) | FishNet-150 · ImageNet-1k val | Top-1 error: 22.51% -> 21.93% | Improved |
| Use 2×2 max-pooling instead of 2×2 average-pooling (ablation) | FishNet-150 · ImageNet-1k val | Top-1 error: 22.86% -> 21.93% | Improved |
| Use 2×2 max-pooling instead of stride-2 convolution (ablation) | FishNet-150 · ImageNet-1k val | Top-1 error: 22.75% -> 21.93% | Improved |
Later papers that compared with this method
1 later accepted paper reports FishNet150 as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Aggregation With Feature Detection (ICCV 2021 · better on 3 of 3 shared results)
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
- Res2Net: A New Multi-Scale Backbone Architecture (TPAMI 2019 · similarity 0.86)
- Fully Convolutional Networks for Semantic Segmentation (TPAMI 2016 · similarity 0.86)
- FishEyeRecNet: A Multi-Context Collaborative Deep Network for Fisheye Image Rectification (ECCV 2018 · similarity 0.86)
- Structured Convolutions for Efficient Neural Network Design (NeurIPS 2020 · similarity 0.85)
- Battle of the Backbones: A Large-Scale Comparison of Pretrained Models across Computer Vision Tasks (NeurIPS 2023 · similarity 0.85)
- LSNet: See Large, Focus Small (CVPR 2025 · similarity 0.84)
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.
