AI Paper Atlas

Research map / Detection and segmentation / CNN features

Rethinking ImageNet Pre-Training

Kaiming He, Ross Girshick, Piotr Dollar

ICCV 2019 · Detection and segmentation › CNN features · approach: Convolution-transformer backbones · 1,214 citations (OpenAlex)

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 changedWhereMeasuredEffect
add ImageNet pre-training (ablation)Mask R-CNN R50-FPN GN, 2x schedule · COCO val2017AP bbox: 36.8 -> 40.3Improved
add ImageNet pre-training (ablation)Mask R-CNN R50-FPN GN, 6x schedule · COCO val2017AP bbox: 41.3 -> 41.1No clear effect
add ImageNet pre-training (ablation)Mask R-CNN R101-FPN GN, 6x schedule · COCO val2017AP bbox: 42.7 -> 42.2No clear effect
train from scratch for 6x instead of 2x iterations (ablation)Mask R-CNN R50-FPN GN · COCO val2017AP bbox: 36.8 -> 41.3Improved
add ImageNet pre-training at AP75 (R50) (ablation)Mask R-CNN R50-FPN GN best schedule · COCO val2017AP bbox 75: 44.6 -> 45.6Hurt
add ImageNet pre-training to ResNeXt-152 (ablation)Mask R-CNN X152 GN · COCO val2017AP bbox / AP mask: 50.9/43.2 -> 50.3/42.5No clear effect
add ImageNet pre-training with test-time augmentation (ablation)Mask R-CNN R50 · COCO val2017AP mask: 41.6 -> 40.9No clear effect
add ImageNet pre-training for keypoint detection (ablation)Mask R-CNN R50-FPN GN · COCO keypoints val2017keypoint AP: 65.6 -> 65.5No clear effect
add ImageNet pre-training with 35k COCO images (ablation)Mask R-CNN R50-FPN GN · COCO 35k train imagesAP bbox: 36.3 -> 36.3No clear effect
add ImageNet pre-training with 10k COCO images (ablation)Mask R-CNN R50-FPN GN · COCO 10k train imagesAP bbox: 25.9 -> 26.0No clear effect
add ImageNet pre-training with 3.5k COCO images (ablation)Mask R-CNN R50-FPN GN · COCO 3.5k train imagesAP bbox: 9.3 -> 16.0Improved
add ImageNet pre-training with 1k COCO images (ablation)Mask R-CNN R50-FPN GN · COCO 1k train imagesAP bbox: 3.5 -> 9.9Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy