Video Object Detection on ImageNetVID
80PrecisionMR2-ByteTrack
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MR2-ByteTrackObject detector Backbone=YOLOXS, Params [M]=9.0, GMAC=10.9, Δ params=0, Δ MAC=02026.05 | 80 | 65.9 | 72.5 | 62.4 | 2.6 | |
| YOLOV-SObject detector Backbone=YOLOXS, Params [M]=10.3, GMAC=12.9, Δ params=+14%, Δ MAC=+21%2026.05 | 79.9 | 66.4 | 72.5 | 62.5 | 2.7 | |
| MR2-ByteTrackObject detector Backbone=YOLOXS, P=2, Params [M]=9.0, GMAC=6.2, Δ params=0, Δ MAC=-43%2026.05 | 77.9 | 63.2 | 69.9 | 57.9 | -1.1 | |
| MR2-ByteTrackObject detector Backbone=NanoDet^3 [17], P=5, Params [M]=1.2, GMAC=0.2, Δ params=0, Δ MAC=-53%2026.05 | 71.2 | 54.1 | 61.5 | — | — | |
| Liu et. alObject detector Backbone=SSDLite-Mobilenetv2 [35], Params [M]=4.9, GMAC=0.2, Δ params=+11%^1, Δ MAC=-84%2026.05 | — | — | — | 61.4 | 0.9 |