Research map / Architectures and efficiency / RNNs and normalization
SemifreddoNets: Partially Frozen Neural Networks for Efficient Computer Vision Systems
Leo F Isikdogan, Bhavin V Nayak, Chyuan-Tyng Wu, Joao Peralta Moreira, Sushma Rao, Gilad Michael
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 |
|---|---|---|---|
| freeze a larger portion of backbone weights (0%, 50%, 75%, 100%) (ablation) | 2x-wide ImageNet-pretrained backbone · Cityscapes semantic segmentation | pixelwise accuracy | Hurt |
| repeat the last Semifreddo blocks twice (ablation) | SemifreddoNet · face identification | accuracy: 88% -> 96% | Improved |
| repeat the blocks further (ablation) | SemifreddoNet · face identification | accuracy: up to 2% | No clear effect |
| add cross-core channel shuffle between the two trainable cores (ablation) | SemifreddoNet · classification and face tasks | performance | Improved |
| keep the first pointwise convolution in ShuffleNetV2 blocks (instead of dropping it) (ablation) | SemifreddoNet · face identification | accuracy | No clear effect |
| keep the first pointwise convolution in ShuffleNetV2 blocks (instead of dropping it) (ablation) | SemifreddoNet · ImageNet classification | top-5 accuracy: 2% absolute | Improved |
| fine-tune batch norm parameters in the frozen core (ablation) | SemifreddoNet · face identification | accuracy: 2% absolute | Improved |
Closest papers by meaning
- SmartFRZ: An Efficient Training Framework using Attention-Based Layer Freezing (ICLR 2023 · similarity 0.84)
- Layerwise Progressive Freezing: A Training Scaffold for Depth-Scalable Binary Networks (NeurIPS 2026 · similarity 0.83)
- Explicitly Modeling Subcortical Vision with a Neuro-Inspired Front-End Improves CNN Robustness (NeurIPS 2025 · similarity 0.83)
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (NeurIPS 2024 · similarity 0.83)
- LSNet: See Large, Focus Small (CVPR 2025 · similarity 0.83)
- Preserving Deep Representations in One-Shot Pruning: A Hessian-Free Second-Order Optimization Framework (ICLR 2025 · similarity 0.83)
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.
