Efficient Two-Stage Detection of Human-Object Interactions with a Novel Unary-Pairwise Transformer

Benchmark Model Rank Results
human-object-interaction-detection-on-hico-detUPT-R101-DC5#20mAP: 32.62Time Per Frame (ms): 124
human-object-interaction-detection-on-hico-detUPT-R101#21mAP: 32.31Time Per Frame (ms): 61
human-object-interaction-detection-on-hico-detUPT-R50#24mAP: 31.66Time Per Frame (ms): 42
human-object-interaction-detection-on-v-cocoUPT-R101-DC5#10AP(S1): 61.3AP(S2): 67.1Time Per Frame(ms): 131
human-object-interaction-detection-on-v-cocoUPT-R101#11AP(S1): 60.7AP(S2): 66.2Time Per Frame(ms): 64
human-object-interaction-detection-on-v-cocoUPT-R50#12AP(S1): 59.0AP(S2): 64.5Time Per Frame(ms): 43