Issue with Camera's GetRGB() Function in Docker Submission

[TomatoPicker278]

In my code, the GetRGB() function of the camera is used, which requires adding the --graphics option when running run_python.shell with the command python my_test_bathing.py --graphics. This is essential for rendering the 3D animation; otherwise, the output will only be a blank image. I have tested this on my local machine, and the results are as shown in the attached image.

However, the submitted Docker does not support the --graphics option. This means that my algorithm cannot be successfully executed within the Docker environment, even though it runs smoothly with graphics acceleration on my local machine.

In my local machine, GetRGB() using "--graphics" is below:

In my local machine, GetRGB() using nothing is below:

[YoruCathy]

Let me see if I can configure the server to enable it to run in headless mode

[TomatoPicker278]

THANKS！ I'm waiting for good news!

[TomatoPicker278]

Below is my test code, the file name is "test_bathing_camera.py", located in ~/RCareWorld/template/, you can try to run this file on your local computer.

`from pyrcareworld.envs.bathing_env import BathingEnv
import numpy as np
import cv2
import argparse

def _main(use_graphics=False):
if use_graphics:
print("using graphics!")
# Initialize the environment
env = BathingEnv(graphics=use_graphics)

env.step(50)

nd_main_intrinsic_matrix = np.array([[600, 0, 240],
                                     [0, 600, 240],
                                     [0, 0, 1]])
nd_ir_intrinsic_matrix = np.array([[480, 0, 240],
                                   [0, 480, 240],
                                   [0, 0, 1]])

camera = env.get_camera()
env.step()

camera.SetTransform([0.11-0.6, 1.00+0.2, 0.0+0.6], [20.0, 130.0, 0.0])
env.step()
env.AlignCamera(654321)
env.step()
camera.GetRGB(512, 512)
env.step()
rgb = np.frombuffer(camera.data["rgb"], dtype=np.uint8)
rgb = cv2.imdecode(rgb, cv2.IMREAD_COLOR)
cv2.imwrite("test_bathing_camera_RGB_1.png", rgb)

if name == "main":
parser = argparse.ArgumentParser(description='Run RCareWorld bathing environment simulation.')
parser.add_argument('-g', '--graphics', action='store_true', help='Enable graphics')
args = parser.parse_args()
_main(use_graphics=args.graphics)`

(1) If it is with '-g' or '--graphics', the result is as shown below.

(2) Otherwise, it is as shown below.

However, if it is run under docker, only the second case (2)

[shxingch]

Hi @TomatoPicker278! We've conducted some tests and found that the provided docker container doesn't include a graphical interface. Since running a virtual graphical interface requires significant computational resources, we've decided to modify our approach.

For future testing, we'll transfer the algorithms from the uploaded docker to our evaluation pc locally, which will help avoid these camera issues. When submitting your work, you only need to ensure that the algorithm modules are complete and properly included.

Regarding your previous submission, we've rerun the tests and can confirm that it now runs without any errors. The relevant logs and videos have been sent to your email.

[TomatoPicker278]

thanks!

[YoruCathy]

Made an issue post documenting it here #180