npu-yanchi / diff-gaussian-rasterization-for-gsslam Goto Github PK
View Code? Open in Web Editor NEWLicense: Other
License: Other
@NPU-YanChi Thank you for your excellent work. What is the difference between tracking and mapping mode? Will it affect the gradient of the rendered color map and the rendered depth?
Hello!
Just tried your implementation with pose backward estimation backward and now im facing some troubles with gradient propagation - cam_q_w2c and cam_t_w2c values has none grad after backward.
Btw i create them from numpy with cast to float 32 and using cuda device with requires_grad=True
cam_q_w2c = torch.tensor(Rotation.from_matrix(R).as_quat(), device='cuda:0', requires_grad=True).float()
cam_t_w2c = torch.tensor(T, device='cuda', requires_grad=True).float()
cam_info = CameraInfo(uid=i, R=R, T=T, cam_q_w2c=cam_q_w2c, cam_t_w2c=cam_t_w2c , FovY=FovY, FovX=FovX, cx=cx, cy=cy, fx=K_train, fy=K_train, image=image, image_path=image_path, image_name=image_name, width=image.size[0], height=image.size[1])
Rasterizer settings and function calling looks like this:
raster_settings = GaussianRasterizationSettings(
image_height=int(viewpoint_camera.image_height),
image_width=int(viewpoint_camera.image_width),
tanfovx=tanfovx,
tanfovy=tanfovy,
bg=bg_color,
scale_modifier=scaling_modifier,
viewmatrix=viewmatrix,
projmatrix=viewpoint_camera.full_proj_transform,
sh_degree=pc.active_sh_degree,
campos=viewpoint_camera.camera_center,
prefiltered=False,
mapping_mode=False,
tracking_mode=True,
debug=False
)
# Rasterize visible Gaussians to image, obtain their radii (on screen).
rendered_image, radii, rendered_depth, rendered_alpha, render_depth_var = rasterizer(
means3D = means3D,
means2D = means2D,
shs = shs,
colors_precomp = colors_precomp,
opacities = opacity,
scales = scales,
rotations = rotations,
cov3D_precomp = cov3D_precomp,
cam_q_w2c = viewpoint_camera.cam_q_w2c,
cam_t_w2c = viewpoint_camera.cam_t_w2c)
Am I doing something wrong?
Thank you.
Hi authors,
I have a minor question on the "Bundle Adjustment" in the paper.
Did you conduct the ablation study on bundle adjustment? I ever tried some simple experiments, which show the bundle adjustment makes the result worse.
I wonder did you do experiments on it? Do you have some observations during the process?
Best,
Shuo.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.