Getting Image data

[Meet001]

I am trying to train a model based on the vision data, I have to set Display setting to 64x64 for doing so, on doing that I am able to get the correct vision data, but now the game runs for training and testing at 64x64, so I am unable to visually see how my car is learning or performing (the only way to know good performance, in this case, is to see if reward is increasing as it should epoch after epoch). on setting size > 64x64 the vision output becomes incorrect. So, is there a way to get correct 64x64 data in python but run torcs on higher resolution?

[xkxiong]

The 64x64 image I extracted is completely black, do you know what the reason is?

[YFrisch]

The 64x64 image I extracted is completely black, do you know what the reason is?

Hey I got the same problem. Could you solve it? I could track it down to the snakeoil client.
client.S.d['img'] is already a vector of 0s.

Edit: Seems to depend for me on using a SSH connection to the remote machine where I run my code. If I run it locally, it works just fine.

[voiler]

The 64x64 image I extracted is completely black, do you know what the reason is?

did you solve this problem? I got the same problem.

[YFrisch]

The 64x64 image I extracted is completely black, do you know what the reason is?

did you solve this problem? I got the same problem.

No. I worked directly on the machine from that on