search

visual-layer / fastdup

fastdup is a powerful free tool designed to rapidly extract valuable insights from your image & video datasets. Assisting you to increase your dataset images & labels quality and reduce your data operations costs at an unparalleled scale.
Other
1.54k stars 74 forks source link

[Feature Request]: COCO formate read json #202

Closed Lifeguard-alex closed 1 year ago

Lifeguard-alex commented 1 year ago

Feature Name

coco formate

Feature Description

i dont see option to read coco formate json file for object detector

Contact Information [Optional]

alex@lifeguard-ai.com

amirmk89 commented 1 year ago

Duplicate of #203

Lifeguard-alex commented 1 year ago

hi , thnks for the answer the code exmaple is missing this : defaultdict(int) class_id img_id

do you have a full code exmaple for this ?

dbickson commented 1 year ago

hi @Lifeguard-alex send us your json and we will make an example for you it seems your json does not include img_id and class_id which are supposed to be there.

dbickson commented 1 year ago

hi @Lifeguard-alex coco annotations are now supported in fastdup, see this Kaggle tutorial notebook: https://www.kaggle.com/code/graphlab/fastdup-coco-format-tutorial