uuberoy / scalabel

Quantify computer vision performance in human terms

---

Scalabel (pronounced "scalable") is a versatile and scalable tool that supports all kinds of annotations needed in a driving database. It supports bounding box, semantic instance segmentation, and video tracking.

Setup

1. Checkout the code

   git clone [email protected]:ucbdrive/scalabel.git
   cd scalabel
   

2. Create a directory <data_dir> to store the server data:

   mkdir ../data
   

3. Launch server. There are two options, either (i) to build with Docker or (ii) to build by yourself.

   1. Build and run a Docker image from the Dockerfile.

      Build by yourself

      docker build . -t 'scalabel/server'
      

      Or

      docker pull scalabel/server
      

      After getting the docker image, you can run the server

      docker run -it -v `pwd`/../data:/go/data -p 8686:8686 scalabel/server
      

   2. Build the server by yourself.

      1. Install GoLang. Refer to the instruction page for details.
      2. Install GoLang dependency

      go get gopkg.in/yaml.v2
      

      3. Compile the packages

      go build -i -o bin/scalabel ./server/go
      

      4. Specify basic configurations (e.g. the port to start the server, the data directory to store the image annotations, etc) in your own config.yml. Refer to app/config/default_config.yml for the default configurations.
      5. Launch the server by running

      ./bin/scalabel --config app/config/default_config.yml
      

4. Access the server through the specifed port (we use 8686 as the default port specified in the config.yml)

   http://localhost:8686
   

Usage and Demo

Please check the documentation for detailed usage instructions.