negative samples

[hdm30]

Hi,

I've tried to add negative samples to the training data by adding the images and leaving the labels (.txt files) empty, However, i encountered the following errors:
File "/home/XXXX/yolov7-segmentation/utils/plots.py", line 90, in box_label
AttributeError: 'FreeTypeFont' object has no attribute 'getsize'
and
File "/home/XXXX/yolov7-segmentation/utils/segment/augmentations.py", line 21, in mixup
segments = np.concatenate((segments, segments2), 0)
ValueError: all the input arrays must have same number of dimensions, but the array at index 0 has 1 dimension(s) and the array at index 1 has 3 dimension(s)

I would like to ask whether your code supports adding negative samples.
Many thanks in advance.

[ortizTorresJJ]

Did you already solve this issue? I have read about other repositories and they do it by adding images without the label info. Nevertheless, this does not work for this repository but I wonder if it is any way to add negative samples

[hdm30]

i solve it by disabling mixup augmentation

[ortizTorresJJ]

@hdm30 And how did you setup the mixup augmentation? in the hyps? or in the arguments in train.py? also did you add images without the label info? or did you add a void .txt?. Thanks in advance!!

[hdm30]

i've just commented the line of mixup augmentation i can't remember where exactly but you could find it easily. yes i've added void txt files