Simulation·Python·MIT
TACTO
Fast, photo-real simulator for vision-based tactile sensors.
Overview
TACTO renders high-resolution tactile readings for sensors like DIGIT and OmniTact directly inside a PyBullet physics scene. It closes the sim-to-real gap for touch by producing the same gel-deformation images your real sensor sees, so you can train contact-rich policies without wearing out hardware.
Install
pip
pip install tactofrom source
git clone https://github.com/facebookresearch/tacto && cd tacto && pip install -e .Example · Render a tactile frame in simulation
example.py
import tacto
import pybullet as p
# Attach a DIGIT sensor model to a simulated gripper
digit = tacto.Sensor(width=240, height=320, visualize_gui=True)
p.connect(p.DIRECT)
digit.add_camera(bodyID, linkIDs=[-1])
# Load an object to touch and read the tactile frame
obj = digit.add_object("objects/cube.urdf")
color, depth = digit.render() # color = gel image, depth = deformation
digit.updateGUI(color, depth)