Jianyuan Wang comments

Results 238 comments of


                                            Jianyuan Wang

Training code release

Closing this issue since the complete training code is now available.

Trajectories txt files

Hi may I ask is there any update for camera traj? Thanks!

make load_and_preprocess_images accept grayscale image

Hi, Since our model expects 3-channel input images, I believe that even grayscale inputs would need to be converted to a 3-channel format (e.g., by converting to RGB or duplicating...

Scene normalization details during training

Hi, thanks! In general, we clamp excessively large depth values to zero by setting a maximum threshold—see this line for reference: https://github.com/facebookresearch/vggt/blob/3d0427aa51af36680b3bec9aeb30a1b5a812893a/training/data/dataset_util.py#L260 Pixels with depth = 0 are subsequently ignored...

Scene normalization details during training

I thought we used: ``` self.depth_max = 24 depth_map = threshold_depth_map(depth_map, max_percentile=-1, min_percentile=-1, max_depth = self.depth_max) ``` For real-world outdoor datasets, usually the points >80m are unreliable (typical for Lidar)

Integrating Viser with Hugging Face Spaces: Handling Port Restrictions

Thanks a lot! I am going to check it in details :)

add new camera to existing sparse scene

Could you elaborate more about "add this image to the original sparse scene"? Do you mean adding this image to the colmap reconstruction output, e.g., images.bin, cameras.bin, and points3D.bin? If...

Call for suggestions on novel view synthesis(NVS) downstream task

Hi, Yes, I believe it’s possible. For fine-tuning, you can use photometric loss and camera pose loss, based on multi-view consistency. For example, given several images, you can first predict...

Code License?

Hi the answer would be yes. I am going through the official process for this, and should be able to update the code license within this month.

Code License?

Closing since We have relicensed to permit commercial use now.