colmap / glomap

GLOMAP - Global Structured-from-Motion Revisited
BSD 3-Clause "New" or "Revised" License
1.3k stars 76 forks source link

The point cloud effect is not good after mapper #86

Open liuxj1 opened 2 weeks ago

liuxj1 commented 2 weeks ago

Hello! Recently, I found in my experiments that the results of some videos when using the mapper are not good, and the sparse point clouds obtained often show a clumping structure? What is the reason for this? Looking forward to your answer

lpanaf commented 2 weeks ago

Hi, thanks for the feedback. If possible, can you maybe share some of the datasets and results so we can better understand the problem?

liuxj1 commented 1 week ago

Hi, thanks for the feedback. If possible, can you maybe share some of the datasets and results so we can better understand the problem?

log: https://drive.google.com/file/d/1HJhPeARv5s9zyAmuDNh1KKxvlDg_-yuL/view?usp=drive_link images: https://drive.google.com/file/d/1l3rzgN2Kjyv7cgLSfAr6Bool90FPmxC1/view?usp=drive_link The point cloud obtained is a cube. Why is this?

HungNgoCT commented 58 minutes ago

"The point cloud obtained is a cube. Why is this?" I have the same problem. I used colmap: feature extractor, exhaustive matcher, mapper: Result was good. but feature extractor, exhaustive matcher, GLOmap: point cloud and camera posed arranged in a cubic.