Research map / Architectures and efficiency / CNN design
Towards Accurate Binary Convolutional Neural Network
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 |
|---|---|---|---|
| represent weights with 5 binary bases instead of one (ablation) | ABC-Net with full-precision activations on ResNet-18 · ImageNet validation | Top-1 accuracy: 62.8% -> 68.3% | Helped |
| represent weights with 5 binary bases instead of one (ablation) | ABC-Net with full-precision activations on ResNet-18 · ImageNet validation | Top-5 accuracy: 84.4% -> 87.9% | Helped |
| use multiple binary weight bases and activations (M=5, N=5) rather than single binary weight and activation (ablation) | ABC-Net on ResNet-18 · ImageNet validation | Top-1 accuracy: 42.7% -> 65.0% | Helped |
| increase binary activations from 1 to 3 with 3 weight bases (ablation) | ABC-Net on ResNet-18 · ImageNet validation | Top-1 accuracy: 49.1% -> 61.0% | Helped |
| increase binary activations from 3 to 5 with 3 weight bases (ablation) | ABC-Net on ResNet-18 · ImageNet validation | Top-1 accuracy: 61.0% -> 63.1% | Helped |
| increase binary weight bases from 3 to 5 with one activation (ablation) | ABC-Net on ResNet-18 · ImageNet validation | Top-1 accuracy: 49.1% -> 54.1% | Helped |
| increase binary activations from 3 to 5 with 5 weight bases (ablation) | ABC-Net on ResNet-18 · ImageNet validation | Top-1 accuracy: 62.5% -> 65.0% | Helped |
| increase weight bases and activations from 1 to 5 (ablation) | ABC-Net on ResNet-34 · ImageNet validation | Top-1 accuracy: 52.4% -> 68.4% | Helped |
| use 5 binary weight bases and activations instead of full-precision weights and activations | ABC-Net versus full-precision ResNet-18 · ImageNet validation | Top-1 accuracy: 69.3% -> 65.0% | Hurt |
| use 5 binary weight bases and activations instead of full-precision weights and activations | ABC-Net versus full-precision ResNet-18 · ImageNet validation | Top-5 accuracy: 89.2% -> 85.9% | Hurt |
| use 5 binary weight bases and activations rather than a single binary weight and activation | ABC-Net versus XNOR-Net · ImageNet validation | Top-1 accuracy: 51.2% -> 65.0% | Helped |
| use ABC-Net with 5 binary weight bases and activations rather than full-precision weights and activations | CNN on Forest Trails · Forest Trails | classification accuracy: 77.7% -> 78.0% | No clear effect |
Closest papers by meaning
- Training binary neural networks with real-to-binary convolutions (ICLR 2020 · similarity 0.92)
- How to Train a Compact Binary Neural Network with High Accuracy? (AAAI 2017 · similarity 0.89)
- BD-Net: Has Depth-Wise Convolution Ever Been Applied in Binary Neural Networks? (AAAI 2026 · similarity 0.89)
- From Hashing to CNNs: Training Binary Weight Networks via Hashing (AAAI 2018 · similarity 0.89)
- Binarized Neural Networks (NeurIPS 2016 · similarity 0.88)
- Efficient Multi-bit Quantization Network Training via Weight Bias Correction and Bit-wise Coreset Sampling (NeurIPS 2025 · 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.