Research map / Detection and segmentation / CNN features
Spatially Adaptive Computation Time for Residual Networks
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 spatially adaptive computation (SACT) instead of a fixed-depth ResNet-101 | ImageNet classification · ImageNet val, test resolution 352 × 352 | accuracy: 78.37% -> 78.68% | Helped |
| Use spatially adaptive computation rather than the matched non-adaptive baseline | ImageNet classification, SACT τ=0.001 · ImageNet val, test resolution 352 × 352 | accuracy: 78.23% -> 78.68% | Helped |
| Use SACT rather than ACT | ImageNet classification · ImageNet val, test resolution 352 × 352 | accuracy: 77.84% -> 78.68% | Helped |
| Use spatially adaptive computation rather than the matched non-adaptive baseline | ImageNet classification, SACT τ=0.005 · ImageNet val, test resolution 224 × 224 | accuracy: 75.57% -> 75.61% | No clear effect |
| Use SACT τ=0.005 instead of ResNet-50 | Faster R-CNN · COCO val | mAP@[.5, .95]: 25.56 -> 27.61 | Helped |
| Use SACT τ=0.005 rather than ResNet-101 | Faster R-CNN · COCO val | FLOPs (%): 100 -> 56.0 ± 8.5 | Helped |
| Use SACT ponder-cost maps rather than the center baseline | Saliency prediction · cat2000 val | AUC-Judd (%): 83.4 -> 84.7 | Helped |
| Train SACT at 248 × 248 rather than 224 × 224 (ablation) | ImageNet classification, SACT · ImageNet val, test resolution 352 × 352 | accuracy: 78.68% -> 79.16% | Helped |
| Train SACT at 248 × 248 rather than 224 × 224 (ablation) | ImageNet classification, SACT · ImageNet val, test resolution 352 × 352 | FLOPs: 6.6% higher | Hurt |
Closest papers by meaning
- Deep Residual Learning for Image Recognition (CVPR 2016 · similarity 0.88)
- Deep Pyramidal Residual Networks (CVPR 2017 · similarity 0.87)
- Residual Attention Network for Image Classification (CVPR 2017 · similarity 0.87)
- Multi-level Residual Networks from Dynamical Systems View (ICLR 2018 · similarity 0.86)
- Glance and Focus: a Dynamic Approach to Reducing Spatial Redundancy in Image Classification (NeurIPS 2020 · similarity 0.86)
- Dr2Net: Dynamic Reversible Dual-Residual Networks for Memory-Efficient Finetuning (CVPR 2024 · 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.