Matthias Humt

Hi @tom-bu you could have a look at OpenCVs `getOptimalNewCameraMatrix` with `alpha=1` to obtain a new camera matrix which preserves pixels. For reference: 1. https://stackoverflow.com/questions/39432322/what-does-the-getoptimalnewcameramatrix-do-in-opencv 2. https://docs.opencv.org/3.3.0/dc/dbb/tutorial_py_calibration.html (under "Undistortion")

If you have masks for the SfM step, you can apply them to a copy of your images and feed those to the MVS step. In Python this could be...

I've provided an approach in #579 which you can also apply on the estimated depth/normal maps.

Hi @HuangBugWei, I've provided an approach in #579 that might be useful.

Hi @jackson2030, in Python you can use ```python from scipy.spatial.transform import Rotation as R QX, QY, QZ, QW = R.from_euler([roll, pitch, yaw]).as_quat() TX, TY, TZ = x, y, z ```...

Hi, for r2n2, try adding ```python voxels_file = cfg['data']['voxels_file'] if voxels_file is not None: fields['voxels'] = data.VoxelsField(voxels_file) ``` at the end `get_data_fields` (so after line 95) of `occupancy_networks/im2mesh/r2n2/config.py`.

Hi, I've written a small article about this if you are still interested: https://www.hummat.com/learning/2022/02/23/implicit-function-learning.html

Hi, thanks to both of you for making this script and pointing out how to modify it, this is great! I just wanted to quickly share how to modify the...

Probably duplicate of #840

Not sure if this is related but using BlenderProc (or BPY from pip?) in this fashion seems to lead to memory leaks. Memory usage is increasing linearly with each call...