Towards All-in-one Pre-training via Maximizing Multi-modal Mutual Information

Open paper
Benchmark Model Rank Results
object-detection-on-coco-test-devM3I Pre-training (InternImage-H)#3box mAP: 65.4
object-detection-on-lvis-v1-0-minivalM3I Pre-training (InternImage-H, single-scale)#4box AP: 65.8
semantic-segmentation-on-ade20kM3I Pre-training (InternImage-H)#4Validation mIoU: 62.9Params (M): 1310
image-classification-on-imagenetM3I Pre-training (InternImage-H)#18Top 1 Accuracy: 89.6%

Rank counts only results with a code link.