From groggyhimself@gmail.com Wed Jun 12 13:43:22 2024 Date: Wed, 12 Jun 2024 13:43:22 +1000 From: Greg 'groggy' Lehey To: Hugin developers list Subject: Strange Hugin failure A friend of mine recently published a panorama that he had stitched from first principles with Mathematica. It didn't look bad: https://lemis.nyc3.digitaloceanspaces.com/grog/Photos/20240608/small/tapiola-kirma.jpeg Before you go looking at these individual links, I have a summary at the bottom of this message. But I thought it could be done better with Hugin. I was wrong. First, I ran it through my scripts, which effectively run pto_gen, cpfind, celeste_standalone, cpclean and autooptimiser. The result was very uneven: https://lemis.nyc3.digitaloceanspaces.com/grog/Photos/20240608/small/tapiola-optimized.jpeg OK, I thought that maybe I had something in my scripts that wasn't doing the right thing, so I tried running it in a vanilla version of Hugin without any ~/.hugin file. Things were *much* worse. Hugin couldn't align the images at all. It seems that it couldn't understand the exposure info, and it made the component images progressively darker: https://lemis.nyc3.digitaloceanspaces.com/grog/Photos/20240610/small/Tapiola-Hugin.png It also came with a popup "The project covers a big brightness range,". But that's not what the images show. They cover a range 9.8 to 11.4 EV, and that matches the lighting. About the only thing that's unusual is that the photos, taken with a Google Pixel 8 Pro, were taken at a sensitivity of only 15 ISO. But that shouldn't make any difference. I've tried this with the latest version of Hugin and also with a 5 year old one, and the results are the same. You can see the summary, with all the images above and more, at http://www.lemis.com/grog/diary-jun2024.php?topics=p&subtitle=Hugin%20fail&article=D-20240611-005714#D-20240611-005714 , and the images themselves are available at http://www.lemis.com/grog/Day/20240610/tapiola-panorama-photos.zip (about 400 MB). Any ideas?