implementing my own data

[Ador2]

when i run regression.py this error comes up after it says "all done":
Traceback (most recent call last): File "regression_cascade.py", line 467, in <module> model.train(feature_train, label_train.astype(np.int32), responses_train) File "regression_cascade.py", line 177, in train target_in_class[:, chn].astype(np.float32)) cv2.error: OpenCV(4.0.0) /io/opencv/modules/ml/src/svm.cpp:1404: error: (-215:Assertion failed) sv_count != 0 in function 'do_train'

note: i didn't use a camera

[higerra]

The error indicates that the there is no supporting vectors for a specific label. This may be caused by empty categories in the training set. Make sure that you have samples for each class.

[wyknxmtb]

I have met the question, too. I wonder how could I solve this problem? BTW, I use a non-tango phone, the collected position information is empty (pose.txt), but I found in the training process, the position is used to generate responses_train in training_data process. When I try to implement my own data, the responses_train is always being zero. Very appreciate for your help and reply

[higerra]

To be clear, are you collecting training data with a non-tango phone? It won't work unless you have other ways to obtain ground-truth device poses. "pose.txt" will always be generated and preprocessed, with or without Tango, for ease of implementation.

If you train with all zero ground-truth positions, you are likely to trigger this "sv_count != 0" assert.

If you are doing research on this, I recommend you try the following:
#10 (comment)