ViDT: An Efficient and Effective Fully Transformer-based Object Detector

Benchmark Model Rank Results
object-detection-on-coco-val2017ViDT (Swin-base, 50ep)#167box AP: 49.2AP50: 69.4AP75: 53.1APS: 30.6APM: 52.6APL: 66.9
object-detection-on-coco-val2017ViDT (Swin-small, 50ep)#196box AP: 47.5AP50: 67.7AP75: 51.4APS: 29.2APM: 50.7APL: 64.8…
object-detection-on-coco-val2017ViDT (Swin-tiny, 50ep)#238box AP: 44.8AP50: 64.5AP75: 48.7APS: 25.9APM: 47.6APL: 62.1…
object-detection-on-coco-val2017ViDT (Swin-nano, 50ep)#302box AP: 40.4AP50: 59.6AP75: 43.3APS: 23.2APM: 42.5APL: 55.8…