Research map / Detection and segmentation / CNN features
FemtoDet: An Object Detection Baseline for Energy Versus Performance Tradeoffs
Peng Tu, Xu Xie, Guo Ai, Yuexiang Li, Yawen Huang, Yefeng Zheng
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 FPN neck with SharedNeck (ablation) | FemtoDet · PASCAL VOC | AP50: 40.04 -> 42.50 | Improved |
| replace FPN with PAN neck (ablation) | FemtoDet · PASCAL VOC | AP50: 40.04 -> 39.91 | No clear effect |
| replace ReLU with GELU activation (ablation) | MobileNetV2 0.25 · ImageNet | Top1 accuracy / Power: Top1 45.22 -> 47.21, Power 5.04 -> 5.67 | Mixed |
| use 5x5 instead of 3x3 kernel (vanilla conv) (ablation) | ResNet18 · ImageNet | Top1 accuracy / Power: 69.93 -> 71.24 Top1, Power 170.64 -> 198.28 | Mixed |
| use depthwise separable convolution instead of vanilla convolution (ablation) | CNN backbone · ImageNet | Power: 170.64 W (Res18) -> 5.56 W (MBV2 DSC) | Improved |
| add instance boundary enhancement (IBE) module (ablation) | FemtoDet* · PASCAL VOC | AP50: 42.50 -> 45.78 | Improved |
| add recursive warm-restart (RecWR) training (ablation) | FemtoDet* · PASCAL VOC | AP50: 42.50 -> 45.13 | Improved |
| combine IBE and RecWR (ablation) | FemtoDet* · PASCAL VOC | AP50: 42.50 -> 46.31 | Improved |
| train longer (1200 epochs instead of 300) (ablation) | FemtoDet* · PASCAL VOC | AP50: 42.50 -> 42.11 | No clear effect |
| use MixUp in standard training (300 epochs) (ablation) | FemtoDet · PASCAL VOC | AP50: 42.50 -> 36.00 | Hurt |
| include MixUp stage in RecWR (ablation) | FemtoDet · PASCAL VOC | AP50: 45.79 -> 46.31 | Improved |
Closest papers by meaning
- EdgeViTs: Competing Light-Weight CNNs on Mobile Devices with Vision Transformers (ECCV 2022 · similarity 0.86)
- EAutoDet: Efficient Architecture Search for Object Detection (ECCV 2022 · similarity 0.86)
- MobileDets: Searching for Object Detection Architectures for Mobile Accelerators (CVPR 2021 · similarity 0.86)
- FCOS: A Simple and Strong Anchor-free Object Detector (TPAMI 2020 · similarity 0.85)
- MegDet: A Large Mini-Batch Object Detector (CVPR 2018 · similarity 0.85)
- Object Detection using Event Camera: A MoE Heat Conduction based Detector and A New Benchmark Dataset (CVPR 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.
