The easiest way to train a U-NET Image Segmentation model using TensorFlow and labelme
Build U-Net with TensorFlow 2 and train a dataset annotated with labelme.
If you are using macOS, you need to execute the following command before installation.
❯ brew install pyqt
Execute the following command to install the unet environment.
❯ conda create -n unet -y python=3.9 && conda activate unet && pip install -r requirements.txt
Annotate images with labelme.
Save the labeled training data to datasets/train
, and create a new datasets/labels.txt
, the content is the classification names, see https://github.com/wkentaro/labelme/tree/main/examples/semantic_segmentation
Execute the following command to generate the voc dataset.
❯ make voc
Note: If you want to regenerate and overwrite the old one, you can execute the following command.
❯ make re-voc
Train and predict with unet.ipynb