The majority of the code that originates from our research is made available to the community with an open source license. We strive to make our research reproducible, we publish code and experiments related to our publications.
Code is maintained in the group github page: https://github.com/hsp-iit
Some highlights are linked below. If you use our code, do not forget to acknowledge us and cite the corresponding paper (you can find references in the readme of each github repository).
A Dataset for Long-Horizon Language-Grounded Manipulation Tasks in Cluttered Tabletop Environments
Learning grasping with dexterous hands with deep reinforcement learning. Code and papers.
ConCon-Chi: Concept-Context Chimera Benchmark for Personalized Vision-Language Tasks
Sim2Real Bilevel Adaptation for Object Surface Classification using Vision-Based Tactile Sensors
Code for online object detection and instance segmentation.
A multi platform, multi protocol communication middleware.
Dataset of fast moving objects. It contains sequences of YCB objects moving fast. Fast-YCB can be useful to benchmark tracking algorithms, as it was done in the paper ROFT (https://github.com/hsp-iit/roft).
Tracking 6D pose of objects. The algorithm specifically target fast moving objects, by integrating output of DCNN for object segmentation and pose estimation, with optical flow in a kalman framework.
GRASPA is a benchmark that was designed to be reproducible and usable across different robot platforms. We provide printable layout for placing objects, and dockers containing pre-installed software for reproducing our experiments and computing the metrics.
An Instance Segmentation Aided Unscented Kalman Filter for 6D Object Pose and Velocity Tracking