Hello everyone,
I want make a fish detection using haar cascade. I have 183 positive, 270 negative images. Training is completed in a short time. But results are very bad, terrible. I attached positive,negative samples and result.
yea, that's expected. cascades are for rigid things with reliable texture and pose.
that's not, what you have.
you probably tried to train it on "aquarium fish", but that's a far too broad category, and any variation in pose or texture will degrade the detection
I have 183 positive, 270 negative images.
even if it were the appropriate method, you would need like 50x the data
to train something successful,
Thank's your answer. My lectrurer want cascade so i can't use yolo tenserflow etc. Should I have 5 thousand positive 10 thousand negatives? I have a hard time finding the negatives. There is not much underwater photograph without fish. Do you know where I can find more data?
My lectrurer want cascade
that person is an idiot. change school/uni, seriously.
for cascades, you need a lot of positives (cropped), but if your negatives are large enough, not so many (it will take "windows" from that). you can also "black out" the fish from a positive image, and use that as a negative
if you really want(or must) to use cascades, make sure:
Asked: 2020-11-20 16:56:26 -0600
Seen: 352 times
Last updated: Nov 21 '20
