Research map / Detection and segmentation / CNN features
Tied Block Convolution: Leaner and Better CNNs with Shared Thinner Filters
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 |
|---|---|---|---|
| Replace standard convolution with TiedResNet | ResNet50 · ImageNet-1k val | top-1 accuracy: 76.2 -> 77.6 | Helped |
| Replace standard convolution with TiedResNet | ResNet101 · ImageNet-1k val | top-1 accuracy: 77.4 -> 78.8 | Helped |
| Use TiedResNet with the small setting | ResNet50 · Cityscapes val | APmask: 31.5 -> 32.5 | Helped |
| Replace SE fully connected layers with TiedSE at B=8 | SEResNet-50 · ImageNet-1k val | top-1 accuracy: 76.71 -> 77.09 | Helped |
| Replace SE fully connected layers with TiedSE at B=4 | EfficientNet-B0 · ImageNet-1k val | top-1 accuracy: 77.1 -> 77.1 | No clear effect |
| Replace GCB convolution with TiedGCB at B=2 | Mask R-CNN · MS-COCO val-2017 | APbbox: 38.9 -> 39.1 | Helped |
| Use 4 splits with base width 32 rather than 2 splits with base width 48 (ablation) | TiedResNet-50 · ImageNet-1k val | top-1 accuracy: 77.27 -> 77.61 | Helped |
| Use 6 splits with base width 24 rather than 4 splits with base width 32 (ablation) | TiedResNet-50 · ImageNet-1k val | top-1 accuracy: 77.61 -> 77.37 | Hurt |
| Use 8 splits with base width 18 rather than 4 splits with base width 32 (ablation) | TiedResNet-50 · ImageNet-1k val | top-1 accuracy: 77.61 -> 77.21 | Hurt |
| Fuse split outputs with concatenation instead of element-wise sum (ablation) | TiedResNet-50 mixer · ImageNet-1k val | top-1 accuracy: 77.61% -> 77.65% | Helped |
| Use 2 splits with base width 48 rather than 4 splits with base width 32 (ablation) | TiedResNet-50 · ImageNet-1k val | top-1 accuracy: 77.61 -> 77.27 | Hurt |
Closest papers by meaning
- SCNN: A General Distribution Based Statistical Convolutional Neural Network with Application to Video Object Detection (AAAI 2019 · similarity 0.87)
- FSNet: Compression of Deep Convolutional Neural Networks by Filter Summary (ICLR 2020 · similarity 0.86)
- Local Binary Convolutional Neural Networks (CVPR 2017 · similarity 0.86)
- CNN Filter DB: An Empirical Investigation of Trained Convolutional Filters (CVPR 2022 · similarity 0.86)
- Deeply Shared Filter Bases for Parameter-Efficient Convolutional Neural Networks (NeurIPS 2021 · similarity 0.85)
- PICNN: A Pathway towards Interpretable Convolutional Neural Networks (AAAI 2024 · similarity 0.84)
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.