Research map / Detection and segmentation / CNN features
MnasFPN: Learning Latency-Aware Pyramid Architecture for Object Detection on Mobile Devices
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 MnasFPN instead of NAS-FPNLite with MobileNetV2 | MobileNetV2 object detector · COCO test-dev | mAP: 25.1 -> 26.1 | Helped |
| Use MnasFPN instead of NAS-FPNLite with MobileNetV2 | MobileNetV2 object detector · COCO test-dev | latency: 202 -> 183 | Helped |
| Use MobileNetV2 + MnasFPN instead of MobileNetV3 + SSDLite with depth multiplier 0.7 | Mobile object detector · COCO test-dev | mAP: 22.0 -> 23.8 | Helped |
| Use MnasFPN instead of SSDLite with MobileNetV2 | MobileNetV3 object detector · COCO test-dev | mAP: +3.4 mAP | Helped |
| Use MnasFPN instead of SSDLite with MnasNet-A1 | MobileNetV3 object detector · COCO test-dev | mAP: 25.5 vs 23.0 | Helped |
| Add expansion features to the MnasFPN search space (ablation) | MnasFPN detection head · COCO val2017 | mAP | Helped |
| Use the MnasFPN search space instead of the NAS-FPNLite-S search space (ablation) | Latency-aware detector search · COCO val2017 | mAP: MnasFPN at 187 ms is more accurate than NAS-FPNLite-S at 201 ms | Helped |
| Apply latency-aware NAS rather than use the post-hoc NAS-FPNLite design (ablation) | NAS-FPNLite-S detector search · COCO val2017 | mAP | Helped |
| Disable Size Dependent Ordering (ablation) | MnasFPN, 4 repeats · COCO val2017 | mAP: 24.9 -> 24.9 | No clear effect |
| Disable Size Dependent Ordering (ablation) | MnasFPN, 4 repeats · COCO val2017 | latency: 187 -> 195 ms | Hurt |
| Disable Size Dependent Ordering (ablation) | MnasFPN, 5 repeats · COCO val2017 | mAP: 25.5 -> 25.5 | No clear effect |
| Disable Size Dependent Ordering (ablation) | MnasFPN, 5 repeats · COCO val2017 | latency: 196 -> 207 ms | Hurt |
Closest papers by meaning
- MnasNet: Platform-Aware Neural Architecture Search for Mobile (CVPR 2019 · similarity 0.89)
- MobileDets: Searching for Object Detection Architectures for Mobile Accelerators (CVPR 2021 · similarity 0.89)
- NAS-FPN: Learning Scalable Feature Pyramid Architecture for Object Detection (CVPR 2019 · similarity 0.88)
- Mobile Video Object Detection With Temporally-Aware Feature Maps (CVPR 2018 · similarity 0.87)
- SM-NAS: Structural-to-Modular Neural Architecture Search for Object Detection (AAAI 2020 · similarity 0.87)
- SP-NAS: Serial-to-Parallel Backbone Search for Object Detection (CVPR 2020 · similarity 0.86)
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.