Closed Guteentak closed 3 years ago
Did you visually inspect the poses or did you verify in some other way whether the estimated poses are good? If yes, can you check whether the format of the text file that you are submitting matches the one the server expects?
Yes, I checked it through COLMAP gui.
For the text file I submitted, I checked that all test cases consist of only night query cases. For example, it can be seen that it comes out as 'IMG_20161227_173116.jpg 0.0688424 0.83849 -0.059498 -0.537267 -336.085 120.708 651.952'.
You mean you are only submitting poses for nighttime images?
python reconstruction_pipeline_v1_1.py --dataset_path /local/aachen --colmap_path /local/colmap/build/src/exe --method_name sift
Yes, as a result of performing the above code, only the poses of night-time images are estimated. When I checked the provided database_v1_1.db, the images of day-time query were not included, is it affected?
The code provided is for the local feature challenge held at past workshops, which only focuses on the nighttime images. Since no daytime images are included in the submission to the server, reporting 0/0/0 for daytime images is meaningful and expected behavior.
Hello, thank you for revealing a good benchmark.
Is it a normal result to have zero results for day query when submitting results to the benchmark site? It is currently benchmarking using SIFT, and both version 1 and 1.1 of Aachen dataset produce the same results.