Semantic Equitable Clustering: A Simple and Effective Strategy for Clustering Vision Tokens

Benchmark Model Rank Results
object-detection-on-coco-val2017SECViT-B (Cascade Mask R-CNN 3x)–box AP: 55.4AP50: 74.1AP75: 59.9Params (M): 114
object-detection-on-coco-val2017SECViT-B (Mask R-CNN 3x)–box AP: 52.8AP50: 73.6AP75: 57.7Params (M): 75
object-detection-on-coco-val2017SECViT-L (Mask R-CNN 1x)–box AP: 52.0AP50: 73.5AP75: 57.3Params (M): 119
object-detection-on-coco-val2017SECViT-L (RetinaNet 1x)–box AP: 50.2AP50: 71.4AP75: 53.9APS: 33.2APM: 54.5APL: 66.3…