Research map / Architectures and efficiency / CNN design
TBN: Convolutional Neural Network with Ternary Inputs and Binary Weights
Diwen Wan, Fumin Shen, Li Liu, Fan Zhu, Jie Qin, Ling Shao, Heng Tao Shen
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 |
|---|---|---|---|
| use ternary inputs instead of binary inputs (TBN vs XNOR-Network) | AlexNet · ImageNet val top-1 | top-1 accuracy: 44.2 -> 49.7 | Improved |
| use ternary inputs instead of binary inputs | ResNet-18 · ImageNet val top-1 | top-1 accuracy: 51.2 -> 55.6 | Improved |
| use ternary inputs instead of binary inputs | ResNet-34 · ImageNet val top-1 | top-1 accuracy: 55.9 -> 58.2 | Improved |
| use ternary inputs instead of binary inputs | Faster R-CNN ResNet-34 · PASCAL VOC2007 test | mAP: 54.7 -> 59.0 | Improved |
| use ternary inputs instead of binary inputs | SSD 300 ResNet-34 · PASCAL VOC2007 test | mAP: 55.1 -> 59.5 | Improved |
| use ternary inputs instead of binary inputs | VGG-7 · CIFAR-10 | accuracy: 90.02 -> 90.85 | Improved |
| increase ternary threshold delta from 0 to 0.4 (ablation) | TBN Simple Network · CIFAR-10 | accuracy: accuracy increases for delta 0 to 0.4 | Improved |
| increase ternary threshold delta beyond 0.4 (ablation) | TBN Simple Network · CIFAR-10 | error rate: error rate increased quickly | Hurt |
| add non-linear layers (None to PReLU) (ablation) | TBN Simple Network · CIFAR-10 | accuracy: 75.92 -> 81.14 | Improved |
| add PReLU to ResNet-20 (ablation) | TBN ResNet-20 · CIFAR-10 | accuracy: 81.36 -> 84.34 | Improved |
| add Sigmoid to ResNet-20 (ablation) | TBN ResNet-20 · CIFAR-10 | accuracy: 81.36 -> 79.12 | Hurt |
| quantize first and last layers: TBN vs XNOR with fully-precise first/last layers | Simple Network · CIFAR-10 | accuracy: 79.11 -> 81.21 | Improved |
Closest papers by meaning
- XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks (ECCV 2016 · similarity 0.89)
- RTN: Reparameterized Ternary Network (AAAI 2020 · similarity 0.88)
- FATNN: Fast and Accurate Ternary Neural Networks (ICCV 2021 · similarity 0.88)
- Towards Accurate Binary Convolutional Neural Network (NeurIPS 2017 · similarity 0.88)
- BD-Net: Has Depth-Wise Convolution Ever Been Applied in Binary Neural Networks? (AAAI 2026 · similarity 0.87)
- Sub-Bit Neural Networks: Learning To Compress and Accelerate Binary Neural Networks (ICCV 2021 · similarity 0.87)
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.
