AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.75k stars 7.96k forks source link

How can I use parts of MSCOCO dataset to train few classes + my custom class (need python script to convert .json files to many .txt files)) #2089

Open py-lacroix opened 5 years ago

py-lacroix commented 5 years ago

I would like to find a python script which is able to read annotations .json files from COCO dataset in order to convert annotations into YOLO format (many .txt files with bounding boxes like this ). The purpose is to train a YOLOv3 custom model to build a detector with same accuracy as COCO pre-trained models for person, cars and trucks and add a custom class for special trucks which are not detected with pre-trained models. I am sure this python script exists but where?

MingPap commented 5 years ago

convert coco to voc format, then convert voc format to YOLO

py-lacroix commented 5 years ago

Maybe from this? [https://github.com/shaqian/Not-Hotdog/blob/master/yolo/coco2yolo.py]