Research map / Detection and segmentation / CNN features
Deep Residual Learning for Image Recognition
Check 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 residual identity shortcuts to a 34-layer plain network | ImageNet classification · ImageNet 2012 validation | top-1 error: 28.54 -> 25.03 | Helped |
| Increase depth from 18 to 34 layers in a plain network | ImageNet plain network · ImageNet 2012 validation | top-1 error: 27.94 -> 28.54 | Hurt |
| Increase residual-network depth from 18 to 34 layers | ImageNet ResNet · ImageNet 2012 validation | top-1 error: 27.88 -> 25.03 | Helped |
| Use projection shortcuts for dimension changes instead of zero-padding (ablation) | ResNet-34 A · ImageNet 2012 validation | top-1 error: 25.03 -> 24.52 | Helped |
| Use projections for every shortcut instead of only dimension-changing shortcuts (ablation) | ResNet-34 B · ImageNet 2012 validation | top-1 error: 24.52 -> 24.19 | Helped |
| Increase residual-network depth from 20 to 110 layers (ablation) | CIFAR-10 ResNet · CIFAR-10 test | error: 8.75 -> 6.43 | Helped |
| Increase residual-network depth from 110 to 1202 layers (ablation) | CIFAR-10 ResNet · CIFAR-10 test | error: 6.43 -> 7.93 | Hurt |
| Replace VGG-16 with ResNet-101 features | Faster R-CNN · MS COCO validation | mAP@[.5, .95]: 21.2 -> 27.2 | Helped |
| Add box refinement (ablation) | ResNet-101 Faster R-CNN · MS COCO validation | mAP@.5: 48.4 -> 49.9 | Helped |
| Add global context (ablation) | ResNet-101 Faster R-CNN with box refinement · MS COCO validation | mAP@.5: 49.9 -> 51.1 | Helped |
| Add multi-scale testing (ablation) | ResNet-101 Faster R-CNN with box refinement and context · MS COCO validation | mAP@.5: 51.1 -> 53.8 | Helped |
| Add an R-CNN network after the RPN (ablation) | ResNet-101 localization system · ImageNet localization validation | top-5 localization error: 14.4 -> 10.6 | Helped |
Closest papers by meaning
- Residual Attention Network for Image Classification (CVPR 2017 · similarity 0.90)
- Involution: Inverting the Inherence of Convolution for Visual Recognition (CVPR 2021 · similarity 0.90)
- Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning (AAAI 2017 · similarity 0.90)
- Dilated Residual Networks (CVPR 2017 · similarity 0.90)
- ObjectNet: A large-scale bias-controlled dataset for pushing the limits of object recognition models (NeurIPS 2019 · similarity 0.90)
- Revisiting Residual Connections: Orthogonal Updates for Stable and Efficient Deep Networks (NeurIPS 2025 · 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.