Skip to content

Add verbose logging to place, slice and extract commands (fix #4)

Vasily Kuznetsov requested to merge i4 into master

I changed the format of boxes to (X0, Y0: WxH) -- I think it might be a bit confusing that we're using left top corner and not the center (as in YOLO-style annotations) but as long as we know it, it's probably ok. I also added tests and implemented logging for slice subcommand, it looks like this:

Slice

  • source dir: /home/vkuznetsov/Documents/ad-detect/data/ad-versarial-test/combined/
  • target dir: /home/vkuznetsov/Documents/ad-detect/data/ad-versarial-sliced/
  • slicing step: 50 pixels
  • include regions that overlap the slice by at least 50%

cnn.png (1919x1013)

slice (0, 0: 1919x1013)

  • region (459, 63: 1004x242) type 0
  • saved as cnn-0.png

cnn2.png (1919x1013)

slice (0, 0: 1919x1013)

  • region (583, 8: 759x95) type 0
  • region (1220, 660: 312x288) type 0
  • saved as cnn2-0.png

foxnews.png (1919x1013)

slice (0, 0: 1919x1013)

  • region (459, 117: 1004x243) type 0
  • region (1385, 401: 278x146) type 0
  • region (1369, 575: 313x242) type 0
  • saved as foxnews-0.png

foxnews2.png (1919x1013)

slice (0, 0: 1919x1013)

  • region (583, 134: 756x88) type 0
  • region (1283, 564: 315x244) type 0
  • region (862, 968: 313x45) type 0
  • saved as foxnews2-0.png

huffingtonpost.png (1919x1013)

slice (0, 0: 1919x1013)

  • region (580, 567: 760x107) type 0
  • region (1292, 695: 318x265) type 0
  • saved as huffingtonpost-0.png

Merge request reports