https://github.com/michael-lammens/save_yolo_seg_masks

Save segmentation masks from yolos predict.py in <labels> <coordinate> format + GUI that shows masks for manual approval or rejection of predictions
https://github.com/michael-lammens/save_yolo_seg_masks

instance-segmentation labeling-tool machine-learning masking yolov5 yolov7

Last synced: about 2 months ago
JSON representation

Save segmentation masks from yolos predict.py in <labels> <coordinate> format + GUI that shows masks for manual approval or rejection of predictions

• Host: GitHub
• URL: https://github.com/michael-lammens/save_yolo_seg_masks
• Owner: Michael-lammens
• Created: 2024-08-07T04:14:04.000Z (5 months ago)
• Default Branch: main
• Last Pushed: 2024-08-07T04:48:39.000Z (5 months ago)
• Last Synced: 2024-08-07T08:16:53.022Z (5 months ago)
• Topics: instance-segmentation, labeling-tool, machine-learning, masking, yolov5, yolov7
• Language: Python
• Homepage:
• Size: 13.7 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
Input/Inferenced images are 640x640 and upsample=True, otherwise must scale the masks. I havent played with non 640x640 images or tested it but will get around to it eventually

## How It Works:

1. Add --save-masks to segment/predict.py argument

2. During inference, only the masks without the bounded boxes are saved to masks/ in the format of   in a .txt file named as the input image.

  - Polygon coordinates are normalized to image size and saved as their contour coordinates

## To review the masks

in review_predictions/ I made a GUI to visualize and buffer the images in the root directory with their bounded boxes and masks from the original predict.py output

I was too lazy to use the masks from masks/ but will add these + other features eventually

For info on the review_predictions usages I added a readme in there