ashawkey / torch-merf Goto Github PK

related link

Thank you for sharing the wonderful results.

Is there a way to output higher resolution in MeRF renderer?

I've increased the resolution of the density grid, but it has only a slight effect.

Can I increase the cache png image size?

Thank you so much for your work! It helped me a lot!

When I tried to pass cam_near_far in render(), it seems not to work because the parameter is not consequently passed to run().

Another problem is that near_far_from_aabb() in run() is using self.aabb_train during training, which causes in-place operation at the min-max clamper of t values. Should it be a miss or on purpose, as I'm not seeing any reasonable gradients to self.aabb_train?

I tried to draw a PR two fix the above issues.

Hi, I cannot run the web viewer, tried chrome and firefox, both on Windows. I am passing the assets dir (or the json file) as the parameter, and the view stays at

Hi, here is my output with your recommended command. why there is so much noise?

I run the following code:
python main.py data/bonsai/ --workspace trial_merf_360_bonsai --enable_cam_center --eval_cnt 1 --save_cnt 1 --downscale 4 #--fast_baking

And I finally get a weird result like this:

Do you know why this happened and how to fix it?

In paper, It is written that VRAM consumption is way lower than others, 256mb or somethin.
But when I train bicycle dataset with command below, I checked nvidia-smi and it was consuming vram about 1~2GB, and its getting more and more when it goes to next step. It was over 7 GB at the end. Is that other way to check VRAM consumption? or just me having such issue?

~/YBS/torch_merf/torch-merf$ python main.py /data_ssd/YBS/merf_data/bicycle/ --workspace trial_bicycle --enable_cam_center --downscale 4