output .json file seems to lose some frame

[hanningzhang]

I am using colab to process my videos, with demo_inference script. But when I look at the output file, the number jumps. You can see that the images jump from 31 to 48 then to 66. Some of my video have the problem while others not. Could you have a look at it?

{"image_id": "31.jpg", "category_id": 1, "keypoints": [188.22901916503906, 1121.029296875, 0.5595381855964661, 188.22901916503906, 1121.029296875, 0.5862871408462524, 184.5247802734375, 1121.029296875, 0.551311194896698, 199.34169006347656, 1113.620849609375, 0.6588364839553833, 173.412109375, 1117.3250732421875, 0.7185412049293518, 214.15859985351562, 1069.170166015625, 0.5805701017379761, 158.59519958496094, 1072.8743896484375, 0.5466502904891968, 243.79241943359375, 1024.7193603515625, 0.6011742353439331, 147.48251342773438, 1017.3109130859375, 0.7052385210990906, 251.2008819580078, 976.564453125, 0.7762030959129333, 143.77828979492188, 965.4517211914062, 0.8358238935470581, 203.04592895507812, 965.4517211914062, 0.5197396278381348, 177.1163330078125, 965.4517211914062, 0.5196272134780884, 191.93324279785156, 883.958740234375, 0.6233901977539062, 191.93324279785156, 887.6629638671875, 0.5283060669898987, 199.34169006347656, 824.6911010742188, 0.5807968378067017, 203.04592895507812, 824.6911010742188, 0.5625457763671875], "score": 1.7606439590454102, "box": [146.93692016601562, 748.0136108398438, 123.3306884765625, 379.31280517578125], "idx": [0.0]}, {"image_id": "48.jpg", "category_id": 1, "keypoints": [194.9777374267578, 818.625732421875, 0.5936064720153809, 198.25823974609375, 815.3452758789062, 0.6418648362159729, 191.69723510742188, 813.7050170898438, 0.6405532360076904, 204.81924438476562, 815.3452758789062, 0.6066005229949951, 190.05699157714844, 813.7050170898438, 0.4371502697467804, 216.3009796142578, 838.3087768554688, 0.32842788100242615, 186.7764892578125, 841.5892333984375, 0.15920603275299072, 217.9412384033203, 876.0344848632812, 0.3375009000301361, 219.58148193359375, 879.3150024414062, 0.09644301980733871, 242.54498291015625, 849.7904663085938, 0.2742788791656494, 244.18524169921875, 849.7904663085938, 0.08555864542722702, 211.38023376464844, 918.6809692382812, 0.2000928819179535, 185.13624572753906, 913.76025390625, 0.21191777288913727, 208.0997314453125, 856.3515014648438, 0.036751639097929, 194.9777374267578, 851.4307250976562, 0.06118826940655708, 188.41673278808594, 917.040771484375, 0.06570660322904587, 183.49598693847656, 913.76025390625, 0.21173517405986786], "score": 1.2320376634597778, "box": [159.09922790527344, 760.06884765625, 83.23876953125, 167.9615478515625], "idx": [0.0]}, {"image_id": "66.jpg", "category_id": 1,

[Maud-Xia]

It may be that no one appeared in frames 32 to 47 and 49 to 65 of your video, so AlphaPose did not detect the human keypoints information in these frames。