Revision history [back]

Hello everyone,

I believe I've found a bug with the CascadeClassifier, I would like to report it but wanted to check in here first.

Quick description:

1. initialize CascadeClassifier with haarcacscade_profileface.xml
2. call detectMultiScale on roi in image to detect a first face (works as expected)
3. call detectMultiScale on same classifier instance, different roi to detect another face (doesn't work as expected, rectangle is offset to the top right)

Note that if I re-instantiate the classifier and use a fresh instance between step 2 and 3, the classifier works as expected and properly detects the face in step 3. That's what's leading me to believe this is a bug.

Full code used to reproduce the issue is here: https://gist.github.com/paulsc/b91b75a3420f39d0cfa2

Can anyone confirm this is an issue? What are the next steps to report this ?

Thanks,

Paul

Hello everyone,

I believe I've found a bug with the CascadeClassifier, I would like to report it but wanted to check in here first.

Quick description:

1. initialize CascadeClassifier with haarcacscade_profileface.xml
2. call detectMultiScale on roi in image to detect a first face (works as expected)
3. call detectMultiScale on same classifier instance, different roi to detect another face (doesn't work as expected, rectangle is offset to the top right)

Note that if I re-instantiate the classifier and use a fresh instance between step 2 and 3, the classifier works as expected and properly detects the face in step 3. That's what's leading me to believe this is a bug.

Full code used to reproduce the issue is here: https://gist.github.com/paulsc/b91b75a3420f39d0cfa2

bug: C:\fakepath\opencv-bug.jpg

no but (re-initializing classifier): C:\fakepath\opencv-nobug.jpg

orignal file: C:\fakepath\bug-equed.jpg

Can anyone confirm this is an issue? What are the next steps to report this ?

Thanks,

Paul

Hello everyone,

I believe I've found a bug with the CascadeClassifier, I would like to report it but wanted to check in here first.

Quick description:

1. initialize CascadeClassifier with haarcacscade_profileface.xml
2. call detectMultiScale on roi in image to detect a first face (works as expected)
3. call detectMultiScale on same classifier instance, different roi to detect another face (doesn't work as expected, rectangle is offset to the top right)

Note that if I re-instantiate the classifier and use a fresh instance between step 2 and 3, the classifier works as expected and properly detects the face in step 3. That's what's leading me to believe this is a bug.

Full code used to reproduce the issue is here: https://gist.github.com/paulsc/b91b75a3420f39d0cfa2

bug: C:\fakepath\opencv-bug.jpg

no but bug (re-initializing classifier): C:\fakepath\opencv-nobug.jpg

orignal file: C:\fakepath\bug-equed.jpg

Can anyone confirm this is an issue? What are the next steps to report this ?

Thanks,

Paul

Hello everyone,

I believe I've found a bug with the CascadeClassifier, I would like to report it but wanted to check in here first.

I'm running on Mac OSX 10.10.3 and OpenCV 3.0.0. I'm using python.

Quick description:

1. initialize CascadeClassifier with haarcacscade_profileface.xml
2. call detectMultiScale on roi in image to detect a first face (works as expected)
3. call detectMultiScale on same classifier instance, different roi to detect another face (doesn't work as expected, rectangle is offset to the top right)

Note that if I re-instantiate the classifier and use a fresh instance between step 2 and 3, the classifier works as expected and properly detects the face in step 3. That's what's leading me to believe this is a bug.

Full code used to reproduce the issue is here: https://gist.github.com/paulsc/b91b75a3420f39d0cfa2

bug: C:\fakepath\opencv-bug.jpg

no bug (re-initializing classifier): C:\fakepath\opencv-nobug.jpg

orignal file: C:\fakepath\bug-equed.jpg

Can anyone confirm this is an issue? What are the next steps to report this ?

Thanks,

Paul

Hello everyone,

I believe I've found a bug with the CascadeClassifier, I would like to report it but wanted to check in here first.

I'm running on Mac OSX 10.10.3 and OpenCV 3.0.0. I'm using python.

Quick description:

1. initialize CascadeClassifier with haarcacscade_profileface.xml
2. call detectMultiScale on roi in image to detect a first face (works as expected)
3. call detectMultiScale on same classifier instance, different roi to detect another face (doesn't work as expected, rectangle is offset to the top right)

Note that if I re-instantiate the classifier and use a fresh instance between step 2 and 3, the classifier works as expected and properly detects the face in step 3. That's what's leading me to believe this is a bug.

Full code used to reproduce the issue is here: https://gist.github.com/paulsc/b91b75a3420f39d0cfa2

bug: C:\fakepath\opencv-bug.jpg

no bug (re-initializing classifier): C:\fakepath\opencv-nobug.jpg

orignal file: C:\fakepath\bug-equed.jpg

Can anyone confirm this is an issue? What are the next steps to report this ?

Thanks,

Paul

Hello everyone,

I believe I've found a bug with the CascadeClassifier, I would like to report it but wanted to check in here first.

I'm running Opencv 3.0.0 on Mac 10.10.3.

Quick description:

1. initialize CascadeClassifier with haarcacscade_profileface.xml
2. call detectMultiScale on roi in image to detect a first face (works as expected)
3. call detectMultiScale on same classifier instance, different roi to detect another face (doesn't work as expected, rectangle is offset to the top right)

Note that if I re-instantiate the classifier and use a fresh instance between step 2 and 3, the classifier works as expected and properly detects the face in step 3. That's what's leading me to believe this is a bug.

Full code used to reproduce the issue is here: https://gist.github.com/paulsc/b91b75a3420f39d0cfa2

bug:

no bug (re-initializing classifier):

orignal file:

Can anyone confirm this is an issue? What are the next steps to report this ?

Thanks,

Paul