How do you crop your image data automatically?

[dayangkunurfaizah]

Hello dear. If you don't mind, could you please share how do you crop your image mammo automatically? Are there any codes could you share here? Hope that you will reply to my question. Thanks in advance.

[AsmaBaccouche]

Hello, what do you mean by cropping the image mamo? do you mean extracting the ROI images that will be used as input?

[dayangkunurfaizah]

yes, dear. What I mean is that, how do you crop the extracted ROI images as input to the U-Network?

[AsmaBaccouche]

The ROI extraction is done by applying a region detection algorithm like YOLO (check my previous paper/work), and then applying the same coordinates found, on the ground truth mask images. If you don't have that first step (object detection step) or you don't plan to apply it by yourself, you will have to manually prepare the data for the segmentation (annotation and cropping)

[dayangkunurfaizah]

Hi dear, I have read your paperwork. Yes, my first step is also goes to object detection. Could you show me which YOLO algorithm you have applied for object detection? The simple algorithm one to me apply on my data sets

[AsmaBaccouche]

Please check this: https://github.com/AsmaBaccouche/Early-Detection-and-Classification-of-Breast-Cancer/blob/main/yolo3_model.py

[dayangkunurfaizah]

Hello, I found the codes. Thank you so much. But how to link with my folder data sets? Could you please teach where can I link my folder image to the codes? I have downloaded the yolo3_model.py and yolo3_utils.py. Both no issues on my jupyter notebook.

[dayangkunurfaizah]

Could you please tell me, what the first step is before applying YOLO algorithm? Do I need to do a labeling image and create a folder for it?

[AsmaBaccouche]

Also check this main script

• https://github.com/AsmaBaccouche/Early-Detection-and-Classification-of-Breast-Cancer/blob/main/train_yolo.py
• https://github.com/AsmaBaccouche/Early-Detection-and-Classification-of-Breast-Cancer/blob/main/yolo_annotation.py

You need to make small changes to use it on your data, and of course you should prepare folder with the images true coordinates in a txt file for the object detection you want to apply.