mworchel / differentiable-shadow-mapping Goto Github PK

Hi, thanks for your interesting work.

I went through your paper and found that you can also perform light direction estimation task with differentiable shadow mapping.
However, I do not find the corresponding code on this GitHub page.

I simply made the direction differentiable and put it in the optimizer but I got the error "Trying to backward through the graph a second time". Could you please provide me the training code as reported in Table 9? Any suggestion is helpful.

Hi there,
I'm rendering a mesh with a point light source. The camera I use is a perspective camera. The code is shown as follows:

camera = Camera(
    projection_matrix=create_perspective_projection_matrix(fovy=math.atan(T_V_FOV), aspect=w / h, near=0.0, far=15.0, device=device),
    view_matrix=create_lookat_matrix(eye=[0.0, 0.0, 0.0], focus=[0.0, 0.0, 1.0], up=[0.0, 1.0, 0.0]).to(device),
)
scene_mesh = weld_seams(Mesh(scene_pc, scene_face))
l = PointLight(ls_pc[0])
renderer = SimpleRenderer(dr.RasterizeGLContext(device=device))
img = renderer.render([scene_mesh], lights=[l], camera=camera, resolution=(h, w))
print(img.shape)

However, the code stops with the following error:

Is there any possible reasons that may lead to this error?
Thank you.

Thanks for the great work! Would you update the code realse in the near future?

I would be very grateful if you could answer.