PreannotationObjects

Pre annotation of objects project.

1. Repository content:

• IoU informations
• Comparing models
• Label Studio configuration
• Unit tests

2. Files structure

.
├── 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

3. Getting started

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

4. Upload file to Label-Studio

The complete instruction is in Label Studio folder README.

5. Images with annotations

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

6. Short demo:

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!