search

rpautrat / SuperPoint

Efficient neural feature detector and descriptor
MIT License
1.88k stars 416 forks source link

KeyError: 'warped_image' #269

Closed Joshpk29 closed 2 years ago

Joshpk29 commented 2 years ago

Sorry if this ha been asked already but I'm having some issues when I try to run export_descriptor for my custom dataset. I keep getting this error:

Traceback (most recent call last): File "export.py", line 410, in args.func(config, output_dir, args) File "export.py", line 125, in export_descriptor img_0, img_1 = sample["image"], sample["warped_image"] KeyError: 'warped_image'

Has anyone encountered this and knows how to fix it --

rpautrat commented 2 years ago

Hi, does your custom dataset actually implements a "warped_image"?

Joshpk29 commented 2 years ago

I think so -- I essentially copied the COCO.py file and modified the file paths to be my dataset, then in the default config at the top of the file I changed warped_pair to be enabled image

Joshpk29 commented 2 years ago

sorry closed on accident

Joshpk29 commented 2 years ago

I did also generate the patches for my dataset -- so I have a LOL/patches folder with folder 0-15 each has 1.jpg, 2.jpg and H_1_2 file, if i could find a way use these in export that would be good too

Joshpk29 commented 2 years ago

figured out my error in loader.py line 114 change the if statement from just hpatches to hpatches or 'custom_dataset_name'