Research map / Architectures and efficiency / RNNs and normalization
Resolution Adaptive Networks for Efficient Inference
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 resolution adaptation alongside depth adaptation | RANet · CIFAR-10, anytime prediction | top-1 accuracy: +1% | Helped |
| Use resolution adaptation alongside depth adaptation | RANet · CIFAR-100, anytime prediction | top-1 accuracy: 2%–5% | Helped |
| Use resolution adaptation alongside depth adaptation | RANet · ImageNet, anytime prediction | top-1 accuracy: 1%–7% | Helped |
| Use resolution adaptation alongside depth adaptation | RANet · ImageNet, anytime prediction, last classifier | FLOPs: around 27% fewer FLOPs | Helped |
| Use resolution adaptation alongside depth adaptation | RANet · CIFAR-10, budgeted batch classification (<0.2×10^8 FLOPs) | FLOPs at 92% accuracy: 20% fewer FLOPs | Helped |
| Use resolution adaptation alongside depth adaptation | RANet · CIFAR-100, budgeted batch classification | FLOPs at 68% accuracy: about 60% FLOPs | Helped |
| Use resolution adaptation alongside depth adaptation | RANet · CIFAR-10, budgeted batch classification (0.2×10^8 to 0.3×10^8 FLOPs) | top-1 accuracy | No clear effect |
| Use resolution adaptation alongside depth adaptation | RANet · CIFAR-100, budgeted batch classification (>0.2×10^8 FLOPs) | top-1 accuracy: ~0.5% | Helped |
| Use resolution adaptation alongside depth adaptation | RANet · ImageNet, budgeted batch classification | top-1 accuracy: 0.5%, 1% and 1.2% | Helped |
| Use resolution adaptation alongside depth adaptation | RANet · ImageNet, budgeted batch classification at matched accuracy | FLOPs: around 65%, 56% and 44% | Helped |
| Use resolution adaptation alongside depth adaptation | RANet · CIFAR-10 and CIFAR-100, budgeted batch classification | efficiency versus WideResNet: up to 4 times | Helped |
| Process coarse, low-resolution features first and defer finer features to uncertain samples | RANet · CIFAR-10, CIFAR-100, and ImageNet, anytime prediction | top-1 accuracy: null | Helped |
Closest papers by meaning
- Adaptive Neural Networks for Efficient Inference (ICML 2017 · similarity 0.87)
- Dynamic Resolution Network (NeurIPS 2021 · similarity 0.87)
- EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks (ICML 2019 · similarity 0.86)
- Glance and Focus: a Dynamic Approach to Reducing Spatial Redundancy in Image Classification (NeurIPS 2020 · similarity 0.86)
- Deep Back-Projection Networks for Super-Resolution (CVPR 2018 · similarity 0.85)
- Adaptive Depth Networks with Skippable Sub-Paths (NeurIPS 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.