Which image format is the best for haar training? [closed]

asked 2014-07-17 02:43:27 -0600

MaR1oC gravatar image

updated 2020-12-06 05:21:27 -0600


My question might be a dumb one but I wonder if there is an image format that is better for haar cascade training which means that:

  1. The training time is reduced
  2. The results are better

Is there one?

Thanks in advance

PS The University of Utah uses .pgm files for its image databases (e.g. the one for car detection)

edit retag flag offensive reopen merge delete

Closed for the following reason the question is answered, right answer was accepted by sturkmen
close date 2020-12-06 05:20:46.060044



It doesn´t make a difference really. Only thing that influences processing time and quality are resolution. Just 1 thing to keep in mind maybe, if you want little details of your classes trained into the model, then do not go for any image type that has lossy compression of the data.

StevenPuttemans gravatar imageStevenPuttemans ( 2014-07-17 04:21:11 -0600 )edit

I would also add that the OpenCV Mat is store in RAW, therefore for any images (without dealing with compression) you would get the same Mat: loading a PNG or a PPM or a PGM would result to the same image for the training.

Mathieu Barnachon gravatar imageMathieu Barnachon ( 2014-07-17 04:54:53 -0600 )edit