Research map / Detection and segmentation / CNN features
Recurrence along Depth: Deep Convolutional Neural Networks with Recurrent Layer Aggregation
Jingyu Zhao, Yanwen Fang, Guodong Li
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 RLA to ResNet-50 | ResNet-50 · ImageNet validation | Top-1 error (%): 24.70 -> 22.83 | Improved |
| add RLA to ResNet-101 | ResNet-101 · ImageNet validation | Top-1 error (%): 23.60 -> 21.48 | Improved |
| add RLA to ResNet-152 | ResNet-152 · ImageNet validation | Top-1 error (%): 23.00 -> 21.22 | Improved |
| add RLA to ResNet-50 | ResNet-50 with Faster R-CNN · COCO val2017 | AP: 36.4 -> 38.8 | Improved |
| add RLA to ResNet-101 | ResNet-101 with RetinaNet · COCO val2017 | AP: 37.7 -> 40.3 | Improved |
| add RLA to ResNet-50 | ResNet-50 with Mask R-CNN · COCO val2017 | instance segmentation AP: 34.1 -> 35.6 | Improved |
| add RLA to ECANet | ECANet-50 · ImageNet validation | Top-1 error (%): 22.52 -> 22.15 | Improved |
| share 1x1 convolution weights by lag (ablation) | DenseNet · CIFAR-10 | classification error (%): 5.80± 0.33 -> 6.22 | Hurt |
| share 1x1 convolution weights by layer ordinal (ablation) | DenseNet · CIFAR-10 | classification error (%): 5.80± 0.33 -> 6.29 | Hurt |
| increase channels by 32 instead of adding RLA (ablation) | ResNet-50 · ImageNet validation | Top-1 error (%): 24.70 -> 23.54 | Improved |
| remove parameter sharing from RLA (ablation) | RLA-v1 on ResNet-50 · ImageNet validation | Top-1 error (%): 22.83 -> 22.93 | Hurt |
| remove information flow from RLA to the main CNN (ablation) | RLA-v1 on ResNet-50 · ImageNet validation | Top-1 error (%): 22.83 -> 23.42 | Hurt |
Closest papers by meaning
- Deep Layer Aggregation (CVPR 2018 · similarity 0.88)
- Convolutional Neural Networks with Gated Recurrent Connections (TPAMI 2021 · similarity 0.87)
- Learning Implicitly Recurrent CNNs Through Parameter Sharing (ICLR 2019 · similarity 0.87)
- Long-Term Recurrent Convolutional Networks for Visual Recognition and Description (TPAMI 2016 · similarity 0.87)
- The Uncanny Similarity of Recurrence and Depth (ICLR 2022 · similarity 0.87)
- Task-Driven Convolutional Recurrent Models of the Visual System (NeurIPS 2018 · 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.
