Pre annotation of objects project.
.
├── coco <-- images for models experiments
│ ├── annotations
│ ├── images
│ ├── iou
│ └── panoptic_val2017.json
├── iou <-- data for IoU experiments
│ ├── images
│ └── README.md
├── label_studio <-- microservis with example using
│ ├── test_images
│ ├── README.md
│ ├── call_*.py
│ └── server.py
├── models <-- models config and checkpoints files
│ ├── README.md
│ └── experiment_models.json
├── tests <-- unit tests
│ ├── README.md
│ └── test_*.py
├── tools <-- math and auxiliary functions
├── README.md
├── env.sh <-- preparing conda virtual environment
└── requirements.txt
Clone this repo to your computer:
git clone https://github.com/LJaremek/PreannotationObjects.git
cd ./PreannotationObjects/
Please run env.sh to create the conda environment 'preannotation':
chmod +x env.sh
./env.sh
Then, activate the environment:
conda activate preannotation
The complete instruction is in Label Studio folder README.
For reproducing experiments you can download the coco
folder to root project folder from:
https://drive.google.com/drive/folders/1JNP46nw0OVIX_uzomTteNaHThS8vcb_t?usp=sharing
To run local server with endpoints:
python server.py
You can create Label Studio project by execute example file:
python label_studio/call_create_ls_project.py
And then preapre json file for Label Studio project:
python label_studio/call_prepare_ls_json.py
When you have ready json file, you can import it (images with annotations) to Label Studio by:
python label_studio/call_import_images_with_annotations.py
And that is it. You have built a pipeline that creates a Label Studio project, annotates the photos folder, and import the annotated photos into Label Studio. Well done!