Fictionarry
commented
1 month ago HI, we first use COLMAP to estimate the camera poses from the provided images, and then run our method. You can fully refer to the DTU pre-processing to handle your own dataset and then start training following the provided scripts. For recommendation, you can first try to use MVS initialization and run following run_llff_mvs.sh, which brings better performance in many cases but may fail in some also.
Your work is truly impressive! I would greatly appreciate it if you could provide some guidance on how to use my own image data for inference within your framework. I have been attempting to work with my images but could use some help in setting it up correctly. Thank you so much for your assistance!