the annot_id of newest rootnet results(6.10) is incorrect

[Tau-J]

I got the error from dataset.py

bbox = np.array(rootnet_result[str(aid)]['bbox'],dtype=np.float32)

KeyError: '75650'

then I double checked the rootnet_interhand2.6m_output_val.json

I found that

[{'annot_id': 75648, 'bbox': [-41.625, -57.333282470703125, 416.25, 416.25], 'abs_depth': [1198.4647216796875, 1103.005126953125], 'root_valid': [0.0, 0.0]}, {'annot_id': 75649, 'bbox': [279.90289306640625, 24.204498291015625, 100.38922119140625, 100.38922119140625], 'abs_depth': [1224.00341796875, 1226.8614501953125], 'root_valid': [0.0, 0.0]}, {'annot_id': 75651, 'bbox': [-70.4109115600586, 72.7760009765625, 151.77374267578125, 151.77374267578125], 'abs_depth': [1122.958740234375, 1113.6966552734375], 'root_valid': [1.0, 0.0]}]

the annot_id 75650 is missing.

[mks0601]

Are you using 30fps dataset for the training? There is another rootnet result for the 30fps version.

[Tau-J]

Are you using 30fps dataset for the training? There is another rootnet result for the 30fps version.

No, I'm using 5fps dataset. I download your newest rootnet json today. (double-checked again just now)

[mks0601]

[mks0601]

I have no problem. Could you check your setting?

[Tau-J]

I have no problem. Could you check your setting?

Sorry, are you sure that your code is the same as the current master branch?

I cloned this repo and downloaded data again, and still got the error.

[mks0601]

I guess you are using different annotation file. InterHand2.6M_val_data.json does not have annotation with annotation_id == 75650.

[Tau-J]

I guess it's a bug of google driver. The file is correct right now. Thanks for your kind help!