Streamlit app to perform annotation for binary segmentation.
- Use the file picker to load an image to annotate with polygons covering target class pixels
- Save the mask file: this will use the input filename and save to the masks directory
- The generated masks are 8-bit images with 0 representing background and 255 representing target
- The file picker is used to select the image to annotate, if you are working through a long list of images it is easy to loose your place and spend lots of time searching for the next file. A workaround is to periodically move the annotated images to a folder called
annotated, making it easier to find the unannotated images - If you refresh the app page, be sure to update the
Mask directory path:or masks will be saved in the default location (they can be just copied over so not a big deal)
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
streamlit run app.py