Computer Vision Annotation Tool (CVAT)
CVAT is completely re-designed and re-implemented version of Video Annotation Tool from Irvine, California tool. It is free, online, interactive video and image annotation tool for computer vision. It is being used by our team to annotate million of objects with different properties. Many UI and UX decisions are based on feedbacks from professional data annotation team.
- Annotation mode
- Interpolation mode
- Attribute mode
- Segmentation mode
- Tutorial for polygons
After you click the link above:
- Click on "GO TO WORKSPACE" and the CVAT environment will load up
- The environment is backed by a K80 GPU
If you have any questions, please contact Onepanel directly at firstname.lastname@example.org. If you are in the Onepanel application, you can also use the chat icon in the bottom right corner.
Code released under the MIT License.
The instructions below should work for
Ubuntu 16.04. It will probably work on other Operating Systems such as
macOS, but may require minor modifications.
Please read official manual here.
Install docker-compose (1.19.0 or newer)
sudo pip install docker-compose
Build docker images
To build all necessary docker images run
docker-compose build command. By default, in production mode the tool uses PostgreSQL as database, Redis for caching.
Run docker containers
To start default container run
docker-compose up -d command. Go to localhost:8080. You should see a login page.
You can include any additional components. Just add corresponding docker-compose file to build or run command:
# Build image with CUDA and OpenVINO support docker-compose -f docker-compose.yml -f components/cuda/docker-compose.cuda.yml -f components/openvino/docker-compose.openvino.yml build # Run containers with CUDA and OpenVINO support docker-compose -f docker-compose.yml -f components/cuda/docker-compose.cuda.yml -f components/openvino/docker-compose.openvino.yml up -d
Additional optional components
- Support for Intel OpenVINO: auto annotation
- Analytics: management and monitoring of data annotation team
- TF Object Detection API: auto annotation
- Support for NVIDIA GPUs
Create superuser account
You can register a user but by default it will not have rights even to view list of tasks. Thus you should create a superuser. The superuser can use admin panel to assign correct groups to the user. Please use the command below:
docker exec -it cvat bash -ic '/usr/bin/python3 ~/manage.py createsuperuser'
Type your login/password for the superuser on the login page and press Login button. Now you should be able to create a new annotation task. Please read documentation for more details.
Stop all containers
The command below will stop and remove containers, networks, volumes, and images
If you want to access you instance of CVAT outside of your localhost you should specify ALLOWED_HOSTS environment variable. The best way to do that is to create docker-compose.override.yml and put all your extra settings here.
version: "2.3" services: cvat: environment: ALLOWED_HOSTS: .example.com ports: - "80:8080"
It is possible to proxy annotation logs from client to ELK. To do that run the following command below:
docker-compose -f docker-compose.yml -f components/analytics/docker-compose.analytics.yml up -d --build
You can use a share storage for data uploading during you are creating a task. To do that you can mount it to CVAT docker container. Example of docker-compose.override.yml for this purpose:
version: "2.3" services: cvat: environment: CVAT_SHARE_URL: "Mounted from /mnt/share host directory" volumes: - cvat_share:/home/django/share:ro volumes: cvat_share: driver_opts: type: none device: /mnt/share o: bind
You can change the share device path to your actual share. For user convenience we have defined the enviroment variable $CVAT_SHARE_URL. This variable contains a text (url for example) which will be being shown in the client-share browser.
CVAT usage related questions or unclear concepts can be posted in our Gitter chat for quick replies from contributors and other users.
However, if you have a feature request or a bug report that can reproduced, feel free to open an issue (with steps to reproduce the bug if it's a bug report).
If you are not sure or just want to browse other users common questions, Gitter chat is the way to go.