TCFormer: Visual Recognition via Token Clustering Transformer

Benchmark Model Rank Results
2d-human-pose-estimation-on-coco-wholebody-1TCFormerV2-Base#15WB: 65.1body: 72.7foot: 74.6face: 78.8hand: 62.6
human-pose-keypoint-estimation-on-coco-test-devRLE + TCFormerV2-Base (384x288)#30AP: 76.1AP50: 92.4AP75: 83.7
human-pose-keypoint-estimation-on-coco-val2017RLE + TCFormerV2-Base (384x288)#34AP: 77.1AP50: 91.0AP75: 83.7AR: 81.5