Research map / Architectures and efficiency / CNN design
Training binary neural networks with real-to-binary convolutions
Read the paperCheck 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 |
|---|---|---|---|
| add the full real-to-binary method to the strong binary baseline | ResNet-18 · ImageNet | Top-1 accuracy: 60.9 -> 65.4 | Helped |
| add attention transfer (ablation) | Strong Baseline ResNet-18 · CIFAR-100, Stage 2, real-valued downsample | Top-1 accuracy: 69.6 -> 72.6 | Helped |
| add HKD alongside attention transfer (ablation) | Strong Baseline ResNet-18 · CIFAR-100, Stage 2, real-valued downsample | Top-1 accuracy: 69.6 -> 73.9 | Helped |
| add gating without attention transfer (ablation) | Strong Baseline ResNet-18 · CIFAR-100, Stage 2, real-valued downsample | Top-1 accuracy: 69.6 -> 70.9 | Helped |
| add progressive teacher-student training (ablation) | Strong Baseline ResNet-18 · CIFAR-100, Stage 2, real-valued downsample | Top-1 accuracy: 69.6 -> 74.6 | Helped |
| combine attention transfer, HKD, and gating in Real-to-Bin (ablation) | ResNet-18 · CIFAR-100, Stage 2, real-valued downsample | Top-1 accuracy: 69.6 -> 76.2 | Helped |
| use gating without attention matching (ablation) | Strong Baseline ResNet-18 · CIFAR-100 | Top-1 accuracy: 69.6 -> 70.9; 72.1 -> 72.0 | No clear effect |
| use progressive teacher-student training instead of ordinary real-valued teacher-student training (ablation) | ResNet-18 · CIFAR-100 | Top-1 accuracy: ~1%; ~5% | Helped |
| add mix-up augmentation (ablation) | Strong Baseline ResNet-18 · CIFAR-100 | Top-1 accuracy | Helped |
| add mix-up augmentation (ablation) | Strong Baseline ResNet-18 · ImageNet | Top-1 accuracy | Hurt |
| use real-valued rather than binary downsample layers (ablation) | Real-to-Bin ResNet-18 · CIFAR-100 | Top-1 accuracy: 5.5%; 6.6% | Mixed |
Closest papers by meaning
- Towards Accurate Binary Convolutional Neural Network (NeurIPS 2017 · similarity 0.92)
- BD-Net: Has Depth-Wise Convolution Ever Been Applied in Binary Neural Networks? (AAAI 2026 · similarity 0.89)
- Local Binary Convolutional Neural Networks (CVPR 2017 · similarity 0.89)
- High-Capacity Expert Binary Networks (ICLR 2021 · similarity 0.89)
- How to Train a Compact Binary Neural Network with High Accuracy? (AAAI 2017 · similarity 0.88)
- Towards Effective Low-Bitwidth Convolutional Neural Networks (CVPR 2018 · similarity 0.88)
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.