Research map / Architectures and efficiency / CNN design
ESPACE: Accelerating Convolutional Neural Networks via Eliminating Spatial and Channel Redundancy
Shaohui Lin, Rongrong Ji, Chao Chen, Feiyue Huang
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 |
|---|---|---|---|
| add spatial-redundancy masking on top of low-rank channel decomposition (ablation) | ESPACE on AlexNet (all conv layers) · ImageNet 2012 val | top-5 error increase at ~4x speedup: LRD 1.37% at 4.15x -> ESPACE 0.01% at 4.15x | Improved |
| use impact mask instead of uniform or random mask (ablation) | ESPACE Conv2 of AlexNet · ImageNet 2012 val | top-5 error increase vs CPU speedup | Improved |
| fine-tune the accelerated network (ablation) | ESPACE masks · AlexNet Conv2 | top-5 error increase | Improved |
| ESPACE vs CPD for whole-network fine-tuning | GoogLeNet · ImageNet 2012 val | top-5 error increase: CPD 3.43% at 3.97x -> ESPACE 0.68% at 4.12x | Improved |
| increase spatial rate reduction from 0.2 to 0.125 (ablation) | ESPACE on AlexNet · ImageNet 2012 val | top-5 error increase: 0.97% -> 8.89% (5.48x -> 8.12x) | Hurt |
| use ESPACE instead of Q-CNN/BC quantization baselines | AlexNet Conv2 · ImageNet 2012 val | speedup at small accuracy loss: 4.72x no loss and 6.11x with 0.1% drop vs Q-CNN 6.06x with 0.33% drop | Improved |
Closest papers by meaning
- Accelerating Very Deep Convolutional Networks for Classification and Detection (TPAMI 2016 · similarity 0.87)
- More Is Less: A More Complicated Network With Less Inference Complexity (CVPR 2017 · similarity 0.87)
- Interspace Pruning: Using Adaptive Filter Representations To Improve Training of Sparse CNNs (CVPR 2022 · similarity 0.87)
- Accelerating Sparse Convolution with Column Vector-Wise Sparsity (NeurIPS 2022 · similarity 0.87)
- Inference, Learning and Attention Mechanisms that Exploit and Preserve Sparsity in CNNs (IJCV 2020 · similarity 0.86)
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (NeurIPS 2024 · 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.
