Sparse R-CNN: End-to-End Object Detection with Learnable Proposals

Benchmark Model Rank Results
2d-object-detection-on-ceymoSparse R-CNN#5mAP: 47.3
2d-object-detection-on-sardet-100kSparse R-CNN#12box mAP: 38.1
object-detection-on-coco-minivalSparse R-CNN (ResNet-101, learnable proposals, random crop aug, FPN)#113box AP: 45.6AP50: 64.6AP75: 49.5APS: 28.3APM: 48.3APL: 61.6
object-detection-on-coco-minivalSparse R-CNN (ResNet-50, learnable proposals, random crop aug, FPN)#129box AP: 44.5AP50: 63.4AP75: 48.2APS: 26.9APM: 47.2APL: 59.5
object-detection-on-coco-minivalSparse R-CNN (ResNet-101, FPN)#139box AP: 43.5AP50: 62.1AP75: 47.2APS: 26.1APM: 46.3APL: 59.7
object-detection-on-coco-minivalSparse R-CNN (ResNet-50, FPN)#155box AP: 42.3AP50: 61.2AP75: 45.7APS: 26.7APM: 44.6APL: 57.6