search

jiexunsee / Simple-Inception-Transfer-Learning

A classifier for cat and dog images (Response to Siraj's challenge of the week)
38 stars 32 forks source link

How to make use of GPU's for this implementation #5

Closed biswasaveek closed 7 years ago

biswasaveek commented 7 years ago

Hi @jiexunsee , I need to get the image vectors of 1000000 images. So, the CPU implementation is too slow for my purpose. I am just starting to dig into how to make use of GPU's for this same code. It will be very helpful if you could provide some insight for that. Thanks in advance.

jiexunsee commented 7 years ago

hi @biswasaveek ,

yes this code takes very long to produce the image vectors, which is why I made another repository here. You can use the producebottlenecks.py script from that repo to get your image vectors fast.

As for GPUs, as long as you have a GPU version of tensorflow installed, you'll be able to run the code using your GPU.

biswasaveek commented 7 years ago

Okay thanks @jiexunsee That works 10x faster than this to give the image vectors. But, the image vector generated for a image in this method and the one you gave reference to is different. Is it supposed to be like that?

jiexunsee commented 7 years ago

Yeahh I think the one here is using a different version of the inception model, so the results would be a little different

On 30 Apr 2017, at 11:39 PM, biswasaveek notifications@github.com wrote:

Okay thanks @jiexunsee That works 10x faster than this to give the image vectors. But, the image vector generated for a image in this method and the one you gave reference to is different. Is it supposed to be like that?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.