So I didn’t find a tool so I could look at my work and double check all the image classifications and I wanted to ensure that I as as accurate as possible. So I wrote one. I used LabelImg to reclassify about 3 categories into 10 categories – and .. sadly after I made this new tool I am glad I did as I am not perfect by any stretch 🙁

Now that I am done – here are my results – is that OK? I don’t now yet

1774 images
100% Complete

Counts
Class 0: aaaa:352
Class 1: bbbb:605
Class 2: occcc:607
Class 3: ddddd:49
Class 4: eeee:50
Class 5: ffff:2
Class 6: gggg:1
Class 7: hhhh:227
Class 8: iiii:263
Class 9: jjjj:187
Class 10: kkkk:169
Class 11: mmmm:39
Class 12: nnnn:38

Total Successful/Valid Extractions: 100%
A reminder that your check folder is: D:\img.check

The number of samples isn’t even, the quality of the original pics was reduced by 4 in the x and y direction? Should I also include the high res images and more .txt files (they can be reused). Not sure! I am going to run my training program for the first time with the best dateset we have yet. Fingers crossed.