Get_camera_image_gpu_tensor: even env get image all inf

Robotics & Edge Computing Robotics - Isaac Isaac Gym
1

Hi.
I’m developing an RL environment with a robot, target object and a camera using isaacGym.
My goal is to make an agent which takes an image as an observation, so I’m using get_camera_image_gpu_tensor and camera attach_camera_to_body. When the env_num is odd, we get right image.However, when the env_num is even, we get all -inf.
I haven’t found the reason. How can I solve this problem？？？

2

Hi @mtchen2016,
I believe using get_camera_image_gpu and attatch_camera_to_body should generate valid images for each environment. I just did a quick test of this with modifying the interop_torch.py example to use attach_camera_to_body, and the images were valid regardless of the number of environments set. If you’d like to share a script that reproduces the issue you are seeing, we can take a look.

3

Hi, I encounter the same question, how did you solve this?