Skip to content

Latest commit

 

History

History
48 lines (30 loc) · 1.29 KB

README.md

File metadata and controls

48 lines (30 loc) · 1.29 KB

tensorflow-unet-labelme

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.

Installation

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

Datasets

Annotate images

Annotate images with labelme.

Generate VOC format dataset

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

Training

Train and predict with unet.ipynb