Research map / Detection and segmentation / CNN features
FNA++: Fast Network Adaptation via Parameter Remapping and Architecture Search
Jiemin Fang, Yuzhu Sun, Qian Zhang, Kangjian Peng, Yuan Li, Wenyu Liu, Xinggang 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 |
|---|---|---|---|
| add FNA++ adaptation to MobileNetV2 | MobileNetV2 seed · Cityscapes DeepLabv3 | mIOU: 75.5 -> 77.0 | Improved |
| add FNA++ adaptation to MobileNetV2 | MobileNetV2 seed · MS-COCO RetinaNet | mAP: 32.8 -> 34.7 | Improved |
| add FNA++ adaptation to MobileNetV2 | MobileNetV2 seed · MPII SimpleBaseline | PCKh@0.5: 85.9 -> 87.0 | Improved |
| add FNA++ adaptation to ResNet-50 | ResNet-50 · MS-COCO RetinaNet | mAP: 35.5 -> 36.8 | Improved |
| add FNA++ adaptation to FBNet-C | FBNet-C · MS-COCO RetinaNet | mAP: 34.9 -> 35.5 | Improved |
| remap parameters from the super network instead of the seed network (ablation) | FNA++ · Cityscapes DeepLabv3 | mIOU: 76.6 -> 77.1 | Improved |
| remap from super network instead of seed network (ablation) | FNA++ · MS-COCO RetinaNet | mAP: 33.7 -> 34.7 | Improved |
| remap from super network instead of seed with 2x schedule (ablation) | FNA++ · MS-COCO RetinaNet 2x | mAP: 35.6 -> 36.0 | Improved |
| remap from super network instead of seed network (ablation) | FNA++ · MS-COCO SSDLite | mAP: 24.0 -> 24.0 | No clear effect |
| add parameter remapping before architecture adaptation (ablation) | FNA++ · Cityscapes DeepLabv3 | mIOU: 76.0 -> 76.6 | Improved |
| add parameter remapping before parameter adaptation (ablation) | FNA++ · Cityscapes DeepLabv3 | mIOU: 73.0 -> 76.6 | Improved |
| replace ImageNet pre-training of target architecture with remapping (ablation) | FNA · Cityscapes DeepLabv3 | mIOU: 76.5 -> 76.6 | 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.
- MobileNetV2 — MobileNetV2: Inverted Residuals and Linear Bottlenecks (CVPR 2018) · better on 4 of 4 shared results
- DetNAS — DetNAS: Backbone Search for Object Detection (NeurIPS 2019) · 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
- Fast Neural Network Adaptation via Parameter Remapping and Architecture Search (ICLR 2020 · similarity 0.97)
- Auto-FPN: Automatic Network Architecture Adaptation for Object Detection Beyond Classification (ICCV 2019 · similarity 0.90)
- DIY Your EasyNAS for Vision: Convolution Operation Merging, Map Channel Reducing, and Search Space to Supernet Conversion Tooling (TPAMI 2023 · similarity 0.88)
- NAS-FCOS: Fast Neural Architecture Search for Object Detection (CVPR 2020 · similarity 0.88)
- Representation Sharing for Fast Object Detector Search and Beyond (ECCV 2020 · similarity 0.88)
- Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation (CVPR 2019 · 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.
