I’ve been attempting to make DeepDream work but have been running into configuration problems on my Macbook. A friend suggested to try it on the cloud, which turned out to be a good idea since I’ve been successful so far in getting images processed by DeepDream following Dockerized instructions from the VISION AI team and a DigitalOcean droplet. If you haven’t created a DigitalOcean account, you can use my referral link to get a $10 credit.
I had to upgrade to the 4GB RAM, 2 CPU droplet just to make it work for 1080p since I was getting out-of-memory errors (error 173). However, it’s still taking roughly 17 minutes for each photo. At 7000 frames, it’s gonna take us 82 days just to finish processing all the frames. So I’m scaling it down to 1280 for now. I’ve signed up to AWS and Google Compute Platform since they offer GPU-enabled instances, but I’ve just learned that you need to request a manual quota increase so you can attach a GPU to your instances. Weird.