TAPTRv3: Spatial and Temporal Context Foster Robust Tracking of Any Point in Long Video

Benchmark Model Rank Results
point-tracking-on-tap-vid-rgb-stacking-query-first-256-256TAPTRv3 (online, ResNet-18)#5AJ: 0.723OA: 0.908delta_avg: 0.841
point-tracking-on-robotap-query-first-256-256TAPTRv3 (online, ResNet-18)#6AJ: 0.645OA: 0.897delta_avg: 0.773
point-tracking-on-tap-vid-davis-query-first-256-256TAPTRv3 (online, ResNet-18)#8AJ: 0.632OA: 0.906delta_avg: 0.764