mantoone

Did you find a solution?

Unfortunately not, I started implementing one, but I am stuck in saving the depth information. I saw some potential solutions here https://stackoverflow.com/questions/47664306/save-depth-images-from-truedepth-camera but I haven't had time to try those...

I was able to write the depth data to a file using Eyal Fink's solution (in the stackoverflow thread). However, now I need to figure out how to read that...

Ok, now I have a working solution that lets you capture depth data and RGB video simultaneously and save those to files. I also created a jupyter notebook for reading...

Did you try training longer? I was able to train Pong-v0 in about 20 minutes with a GTX 1080 Ti, but it took hours with a CPU. By default it...

I've seen similar behavior when I ran out of memory, maybe you need to reduce the number of workers? Training MsPacman-v0 works for me on: Ubuntu 18.04 python 3.6.5 torch...

I guess you can also try to compensate the smaller number of workers with smaller learning rate: ``` python main.py --env MsPacman-v0 --workers 7 --lr 0.00001 2018-10-31 23:13:01,099 : lr:...

Looks like you're not getting an image from the webcam. Do you have a camera available at /dev/video0 ?

Having the same problem. Did you find a solution?