Challenge and Dataset on Large-scale Human-centric Video Analysis in Complex Events (HiEve)


Short Name: RPGNet
Complete name: Relation based Pose Graph Network
Description: An efficient modle for crowded pose estimatation.
Runtime: one week Hz
Submissions: 1

Benchmark
performance:

T_w_AP T_w_AP_50 T_w_AP_75 T_w_AP_90 T_AP T_AP_50 T_AP_75 T_AP_90
47.3351 56.2681 44.6935 41.0437 50.8333 59.6354 48.4024 44.4622

Detailed
performance:


Video Name T_w_AP T_w_AP_50 T_w_AP_75 T_w_AP_90 T_AP T_AP_50 T_AP_75 T_AP_90
hm_in_waiting_hall (ID:20) 54.5308 68.1547 48.5346 46.9031 54.7096 68.2350 48.7703 47.1236
hm_in_bus (ID:21) 33.1979 31.5349 35.7173 32.3414 34.0308 31.8761 37.9919 32.2243
hm_in_dining_room2 (ID:22) 21.7705 27.8126 19.5050 17.9938 21.7793 27.7623 19.5761 17.9994
hm_in_lab2 (ID:23) 66.5569 77.3198 64.5380 57.8129 68.8933 78.6783 67.7510 60.2507
hm_in_subway_station (ID:24) 36.7001 42.6287 35.1361 32.3355 37.3432 43.6078 35.6670 32.7548
hm_in_passage (ID:25) 60.9291 57.2631 63.4922 62.0319 62.1138 58.1095 63.7348 64.4971
hm_in_fighting4 (ID:26) 42.9579 51.0252 42.0370 35.8115 44.0668 50.6835 43.6582 37.8586
hm_in_shopping_mall3 (ID:27) 59.8767 76.6982 56.6134 46.3184 61.2141 77.8119 58.0852 47.7452
hm_in_restaurant (ID:28) 57.6037 62.1406 58.3469 52.3235 55.5188 58.4635 56.0119 52.0810
hm_in_accident (ID:29) 47.7248 55.3489 46.4450 41.3804 47.6385 54.8249 46.4221 41.6686
hm_in_stair3 (ID:30) 44.7780 53.7354 41.6825 38.9160 43.0532 51.8189 39.3501 37.9906
hm_in_crossroad (ID:31) 72.7715 83.7654 72.1337 62.4155 72.9862 83.3625 71.2934 64.3028
hm_in_robbery (ID:32) 45.8156 50.7629 45.2890 41.3950 46.2745 50.9678 45.9279 41.9278