Using Sim2RealViz, the sim2real gap of Data Augmentation model can be compared agaings other models (e.g. Vanilla or Fine-tuned) and displayed on the real-world environment map along with its performance metrics. In particular, Sim2RealViz shows ① that those models are particulary effective in simulation, but we identified errors in the environment, such as the model failing to regress its position because of a closed-door that was opened in training. Such an error can be selected by instance on the map ② to identify key features extracted by the model either as superimposed on the bird's eye-map ③, or as a first person view ④.

For more information, please refer to the manuscript: Sim2RealViz: Visualizing the Sim2Real Gap in Robot Ego-Pose Estimation

Work by: Théo Jaunet, Guillaume Bono, Romain Vuillemot, and Christian Wolf

Step 1: Clone this repo and install Python dependecies as it follows (you may want to use a vitural environment for that).

pip install -r requirements.txt

Step 2: For a direct interaction with models and simulation, this tool requieres both (Habitat-sim + habitat-api) and pytorch

You can follow installation insctructions here:

• habitat
• Pytorch

Step 3: Download the virtual environment data in from this drive, and move it to <project_dir>/data/

  mv ~/Downloads/citi.glb data/citi.glb

Step 4: You can launch the server with the script 'server.py' at the root of this repo.

python server.py

The server should then be accessible at: http://0.0.0.0:5000 (it may take a minute or two to launch).