Error generating tfdata_12net

[vonclites]

After running gen_shuffle_data.py, I get 644,000 negative samples.

https://github.com/wangbm/MTCNN-Tensorflow/blob/4ae3a720dc08aebb107f9ed0335c7b0b5e3e22c6/prepare_data/gen_tfdata_12net.py#L77

generates an error because I cannot sample 1,000,000 times without replacement from a pool of 644,000.

Is this intended to be sampling 100,000 times? Or was gen_shuffle_data.py supposed to generate over a million samples?

Thanks!

[Honourab]

I found the same error with you

[johnny88850tw]

I think that's because the dataset is not enough 1000000, you can set replace=True, but it cause the dataset will be reselected.

[kyawthiha31]

I changed 'replace=True' to my script and still lead to the error.

[galuhputraa]

change line 77 into this:
try:
neg_keep = npr.choice(len(neg), size=1000000, replace=False)
except ValueError:
neg_keep = npr.choice(len(neg), size=1000000, replace=True)