OpenCV and caffe builds compatible with anaconda 4.3.0

[183amir]

I would like to create an environment with anaconda 4.3.0 and some other packages (caffe and opencv) which we would use to develop bob at idiap (https://www.idiap.ch/software/bob/) but as the new version of anaconda has updated its dependencies, the rest of the stack here has not been updated. To be precise, anaconda 4.3.0 depends on jpeg 9 while opencv still uses jpeg 8 and anaconda 4.3.0 depends on mkl 2017.0.1 while caffe depends on mkl 11.3.3.

It would be great if Continuum could release a new version of these packages too which would be compatible with the newest version of anaconda.

I have also opened a pull request here: #80 which would update opencv to use jpeg 9. If needed, I can open another one to update caffe's dependencies.

ping @anjos

[183amir]

There is also rc4 version of caffe released recently https://github.com/BVLC/caffe/releases and opencv too: https://github.com/opencv/opencv/releases

ping @anjos @tiagofrepereira2012

[msarahan]

Thanks, we'll be sure to pick up the jpeg update with the update of both of those packages. I'll try to get to this by the end of the week.

[183amir]

@msarahan is there any news on this? Deep learning is very popular nowadays, I'm sure some other people are waiting for this too :)
I have also another issue open here #74 which is related to deep learning too.

[msarahan]

@183amir sorry, I have been swamped with contract work. I have not gotten to this yet. I'll try to get to this soon. I know how important deep learning is to people, but the bills have to get paid first, so to speak.

[183amir]

Hi, sorry to ping again. Can I do anything to help this be solved? Thanks.

[jjhelmus]

@183amir caffe and caffe-gpu version 1.0.0rc5 packages are available on the defaults channel as well new opencv packages which support numpy 1.12 and mkl 2017.0.1. The opencv package are still targeting jpeg8, let me see if I can push on that issue to get some new builds out. It also looks like some dependencies are missing for Python 3.6, I'll look into that as well.