Add interface for depth in both forward rendering and backward propagation

[ingra14m]

This is the visualization for depth of lego

[grgkopanas]

That's cool thank you, its a duplicate of graphdeco-inria/gaussian-splatting#20 similar things apply, to be able to accept it we either need a test for perfomance in one of the mipnerf360 scenes or tanks and temples (before and after) to make sure there is no regression or a flag that is by default disabling the depth rendering.

edit: by perfomance i mean speed

[JonathonLuiten]

Note this is identical to #2

Even uses the same variable names...
Also, doesn't actually do the backward pass properly?
Probably close as duplicate?

[dongshengLiu]

Hey, How can I display the depth map ?

[ingra14m]

Hi, you can refer to #30 in gaussian-splatting

[dongshengLiu]

Hi, you can refer to #30 in gaussian-splatting

Thanks!

[ThePassedWind]

Hi, I'd like to use your code to get depth image when training (running train.py).
However, an error occurred:
Traceback (most recent call last):
File "D:\PythonWorkplace\gaussian-splatting\train.py", line 227, in training(lp.extract(args), op.extract(args), pp.extract(args), args.test_iterations, args.save_iterations, args.checkpoint_iterations, args.start_checkpoint, args.debug_from)
File "D:\PythonWorkplace\gaussian-splatting\train.py", line 101, in training loss.backward()
File "D:\CodingSoftware\Anaconda\envs\ro_gaussian_splatting\lib\site-packages\torch_tensor.py", line 487, in backward
torch.autograd.backward(
File "D:\CodingSoftware\Anaconda\envs\ro_gaussian_splatting\lib\site-packages\torch\autograd_init_.py", line 200, in backward Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
File "D:\CodingSoftware\Anaconda\envs\ro_gaussian_splatting\lib\site-packages\torch\autograd\function.py", line 275, in apply return user_fn(self, *args)
TypeError: _RasterizeGaussians.backward() takes 3 positional arguments but 4 were given
Training progress: 0%| | 0/30000 [00:00<?, ?it/s]

I have successfully tested getting depth image when running render.py.
Could you please give some suggestions about how to fix this error?

[ingra14m]

Hi, depth backpropagation should not cause an error. I guess you might have used the wrong branch. You can refer to this PR.

The depth branch contains both forward and backward propagation of depth, whereas the main branch only contains the forward propagation of depth.

[ThePassedWind]

Thanks a lot, I found I forgot to convey two parameters in backward function.
The problem is solved!