GLError occurred while running "grounded_sam_osx_demo.py"

[Zasa-Y]

I am running this code in ssh environment without display device. I entered the following code:

export CUDA_VISIBLE_DEVICES=0 python grounded_sam_osx_demo.py \ --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \ --grounded_checkpoint checkpoint/groundingdino_swint_ogc.pth \ --sam_checkpoint checkpoint/sam_vit_h_4b8939.pth \ --osx_checkpoint checkpoint/osx_l_wo_decoder.pth.tar \ --input_image assets/osx/grounded_sam_osx_demo.png \ --output_dir "outputs" \ --box_threshold 0.3 \ --text_threshold 0.25 \ --text_prompt "humans, chairs" \ --device "cuda"

Then I got the following error:

╭─────────────────── Traceback (most recent call last) ─────────────────────╮ │ /workspace/Grounded-Segment-Anything/grounded_sam_osx_demo.py:276 in │ │ 273 │ for i, label in enumerate(pred_phrases): │ 274 │ │ if 'person' in label.lower() or 'man' in label.lower(): │ 275 │ │ │ boxes_human.append(boxes_filt[i]) │ ❱ 276 │ rendered_img = mesh_recovery(image, boxes_human) │ 277 │ cv2.imwrite(os.path.join(output_dir, "grounded_sam_osx_output.jpg"), rendered_img) │ 278 │ │ 279 │ # draw output image │ │ /workspace/Grounded-Segment-Anything/grounded_sam_osx_demo.py:175 in mesh_recovery │ │ 172 │ │ focal = [cfg.focal[0] / cfg.input_body_shape[1] bbox[2], cfg.focal[1] / cfg.in │ 173 │ │ princpt = [cfg.princpt[0] / cfg.input_body_shape[1] bbox[2] + bbox[0], │ 174 │ │ │ │ cfg.princpt[1] / cfg.input_body_shape[0] bbox[3] + bbox[1]] │ ❱ 175 │ │ renderedimg, = render_mesh(vis_img[:, :, ::-1], mesh, smpl_x.face, {'focal': │ 176 │ │ vis_img = rendered_img.copy() │ 177 │ │ 178 │ return rendered_img │ │ /workspace/Grounded-Segment-Anything/grounded-sam-osx/utils/vis.py:152 in render_mesh │ │ 149 │ scene.add(camera) │ 150 │ │ 151 │ # renderer │ ❱ 152 │ renderer = pyrender.OffscreenRenderer(viewport_width=img.shape[1], viewport_height=i │ 153 │ │ 154 │ # light │ 155 │ light = pyrender.DirectionalLight(color=[1.0, 1.0, 1.0], intensity=0.8) │ │ /opt/conda/envs/pytorch-1.12/lib/python3.9/site-packages/pyrender/offscreen.py:31 in init │ │ 28 │ │ │ 29 │ │ self._platform = None │ 30 │ │ self._renderer = None │ ❱ 31 │ │ self._create() │ 32 │ │ 33 │ @property │ 34 │ def viewport_width(self): │ │ /opt/conda/envs/pytorch-1.12/lib/python3.9/site-packages/pyrender/offscreen.py:149 in _create │ │ 146 │ │ │ raise ValueError('Unsupported PyOpenGL platform: {}'.format( │ 147 │ │ │ │ os.environ['PYOPENGL_PLATFORM'] │ 148 │ │ │ )) │ ❱ 149 │ │ self._platform.init_context() │ 150 │ │ self._platform.make_current() │ 151 │ │ self._renderer = Renderer(self.viewport_width, self.viewport_height) │ 152 │ │ /opt/conda/envs/pytorch-1.12/lib/python3.9/site-packages/pyrender/platforms/egl.py:186 in │ init_context │ │ 183 │ │ assert eglBindAPI(EGL_OPENGL_API) │ 184 │ │ │ 185 │ │ # Create an EGL context │ ❱ 186 │ │ self._egl_context = eglCreateContext( │ 187 │ │ │ self._egl_display, configs[0], │ 188 │ │ │ EGL_NO_CONTEXT, context_attributes │ 189 │ │ ) │ │ /opt/conda/envs/pytorch-1.12/lib/python3.9/site-packages/OpenGL/platform/baseplatform.py:402 in │ call │ │ 399 │ │ return None │ 400 │ def call( self, args, *named ): │ 401 │ │ if self.load(): │ ❱ 402 │ │ │ return self( args, **named ) │ 403 │ │ else: │ 404 │ │ │ from OpenGL import error │ 405 │ │ │ raise error.NullFunctionError( │ │ in OpenGL_accelerate.errorchecker._ErrorChecker.glCheckError:58 ╰─────────────────────────────────────────────────────────╯ GLError: GLError( err = 12297, baseOperation = eglCreateContext, cArguments = ( <OpenGL._opaque.EGLDisplay_pointer object at 0x7ff37e08b440>, <OpenGL._opaque.EGLConfig_pointer object at 0x7ff37e08b4c0>, <OpenGL._opaque.EGLContext_pointer object at 0x7ff37df985c0>, <OpenGL.arrays.lists.c_int_Array_7 object at 0x7ff37e08b5c0>, ), result = <OpenGL._opaque.EGLContext_pointer object at 0x7ff37e08bac0> )

Please how can I solve this problem?

[linjing7]

Hi, what's the version of your pyrender and opengl? I also run on a headless device and I do not meet this problem. My setting is: pyrender==0.1.36, pyopengl==3.1.0, python=3.7. I think you can refer to these two issues: issue1 and issue2.

[Zasa-Y]

Thanks! I set pyrender and opengl to exactly version pyrender==0.1.36, pyopengl==3.1.0, but still had the error. Then I tried the solution in issue2, it works. The details are as follows:

Modify [site-packages/pyrender/renderer.py] in line 1059-1067:

glRenderbufferStorageMultisample(
    GL_RENDERBUFFER, 4, GL_RGBA,
    self.viewport_width, self.viewport_height
)
glBindRenderbuffer(GL_RENDERBUFFER, self._main_db_ms)
glRenderbufferStorageMultisample(
    GL_RENDERBUFFER, 4, GL_DEPTH_COMPONENT24,
    self.viewport_width, self.viewport_height
)

change these two 4 to 1.

Then, use this macro in front of the script to run MESA_GL_VERSION_OVERRIDE=4.1 python grounded_sam_osx_demo.py

Problem solved. Thanks again！

Hi, what's the version of your pyrender and opengl? I also run on a headless device and I do not meet this problem. My setting is: pyrender==0.1.36, pyopengl==3.1.0, python=3.7. I think you can refer to these two issues: issue1 and issue2.

[linjing7]

Okay, happy to here this news :)