AgricultureDetection

Detect the crops picture like garlic based on tensorflow-unet and calculate the areas of the garlic seeds.

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.

Project images and datasets are from Okayama University, please replace them with your own datasets when using them

Installation

1.1 If you are using macOS, you need to execute the following command before installation.

> brew install pyqt

1.2 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

2.1 If you are using linux-ubuntu, you can run the Docker to configure your development environment.

> docker build -t image_name:image_version

2.2 Then create your container and go into it.

> docker run -it -p ipaddress:port:binding_port container_name /bin/bash

please replace it with your ipaddress and port.

2.3 Use jupyter notebook to train your datasets.

> jupyter notebook --allow-root

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

Ps: This improved application refers to CatchZeng / tensorflow-unet-labelme