When using the "exhaustive" method to perform matching, COLMAP generates N sparse models, always trying to improve upon the previous result. By default, the original code always selects the best first model ("colmap/sparse/0"), which is not always the best one. The modification proposes a scan of the "colmap/sparse" directory to locate all models, analyze the number of frames found in each one, and then save the "transforms.json" file based on the best model.
When using the "exhaustive" method to perform matching, COLMAP generates N sparse models, always trying to improve upon the previous result. By default, the original code always selects the best first model ("colmap/sparse/0"), which is not always the best one. The modification proposes a scan of the "colmap/sparse" directory to locate all models, analyze the number of frames found in each one, and then save the "transforms.json" file based on the best model.
This PR will resolve some open issues.