yolo dataset format to paddleocr dataset format

[wdcs-krishpatel]

I want to finetune paddleocr. Now I have image and label dataset in a yolov8 format, so how do I convert to padlleocr format and its folder structure and finetune it.

[GreatV]

please refer to:

PaddleOCR/tools/infer/predict_system.py

Lines 124 to 144 in 4336771

	dt_boxes = sorted_boxes(dt_boxes)
	for bno in range(len(dt_boxes)):
	tmp_box = copy.deepcopy(dt_boxes[bno])
	if self.args.det_box_type == "quad":
	img_crop = get_rotate_crop_image(ori_im, tmp_box)
	else:
	img_crop = get_minarea_rect_crop(ori_im, tmp_box)
	img_crop_list.append(img_crop)
	if self.use_angle_cls and cls:
	img_crop_list, angle_list, elapse = self.text_classifier(img_crop_list)
	time_dict["cls"] = elapse
	logger.debug(
	"cls num : {}, elapsed : {}".format(len(img_crop_list), elapse)
	)
	if len(img_crop_list) > 1000:
	logger.debug(
	f"rec crops num: {len(img_crop_list)}, time and memory cost may be large."
	)
	rec_res, elapse = self.text_recognizer(img_crop_list)

[wdcs-krishpatel]

but is this not for inference?

[UserWangZz]

The annotation file format supported by the text detection algorithm in PaddleOCR is as follows, separated by '\t':

" image path                    Image annotation information encoded by json.dumps"
ch4_test_images/img_61.jpg    [{"transcription": "MASA", "points": [[310, 104], [416, 141], [418, 216], [312, 179]]}, {...}]

We do not provide tools for data conversion, but you can easily convert the yolo format to the paddleocr detection format.

If you do not have a field for text content, you can set the transcription field to an empty character, such as "".
Text detection reference document: https://github.com/PaddlePaddle/PaddleOCR/blob/main/doc/doc_en/detection_en.md
Text recognition reference document: https://github.com/PaddlePaddle/PaddleOCR/blob/main/doc/doc_en/recognition_en.md

[wdcs-krishpatel]

Okkay now I got it. Thank you @UserWangZz for the support.