Software

Human-to-Robot Handovers of Unseen Containers with Unknown Filling

Handover Human–Robot Interaction Computer Vision Robot Control Python ROS

System for human-to-robot handover of unseen containers with unknown filling, with perception (vision) and robot control modules.

Towards Safe Human-to-Robot Handovers of Unknown Containers

Robotics Simulation Computer Vision Handover Python PyBullet

Real-to-simulation framework for safe human-to-robot handovers using visual estimation of object and hand properties.

LoDE: Localisation and Object Dimensions Estimator

Computer Vision 3D Reconstruction Multi‑View Python

Joint localisation and dimension estimation of container-like objects using generative 3D sampling and multi-view fitting.

WHC: Whole‑Body Control with QP

Library Robotics Control C++

Generic whole-body control library implementing inverse dynamics and kinematics using quadratic programming.

Carefulness Detection

Robotics Dynamics Classification MATLAB Python ROS

Dynamical systems model that classifies handovers as careful or not based on motion patterns.

ACC: Audio Content Classification

Audio Classification

Models for identifying actions (shaking or pouring) and classifying content type and level using audio spectrograms.

PRIME: A Few Primitives Can Boost Robustness to Common Corruptions

Robustness Data Augmentation Computer Vision

Plug‑and‑play augmentation scheme using max‑entropy transformations to improve robustness to common corruptions.

NADs: Neural Anisotropy Directions

Deep Learning Inductive Bias

Reproduce the results of NADs, capturing directional inductive biases of neural architectures.

Hold Me Tight!

Deep Learning Feature Analysis

Reproduce the analysis of discriminative features and their influence on deep network decision boundaries.

ColorFool: Semantic Adversarial Colorization

Adversarial ML Semantic Segmentation

Generate adversarial images by perturbing semantic regions within natural color ranges.

Toolkit for the CHOC Dataset

Toolkit Pose Estimation Computer Vision

Toolkit with utilities for the CORSMAL Hand-Occluded Containers (CHOC) dataset, including loading, visualization, and evaluation tools.

Toolkit for Rendering Composite Images

Toolkit Rendering Blender

Toolkit to automatically render composite images of handheld containers over real backgrounds using Blender and Python, combining synthetic objects, hands, and forearms.

CHOC-NOCS for 6D Object Pose Estimation with Hand Occlusions

6D Pose Estimation Hand Occlusions Computer Vision

Software to run a multi-branch convolutional neural network, adapted from NOCS and retrained at QMUL, for category-level 6D pose estimation on images of hand-occluded containers.



Software from people using project data and models, software, and benchmarks

Squids
Squids
Improving Generalization of Deep Networks for Estimating Physical Properties of Containers and Fillings
Software of the solution submitted by the team Squids at the 2022 ICASSP CORSMAL challenge.
[code] [paper]
KEIO-ICS Shared Transformer Encoder with Mask-based 3D Model Estimation for Container Mass Estimation
Software of the solution submitted by the team KEIO-ICS at the 2022 ICASSP CORSMAL challenge.
[code]
Visual Container localisation and mass estimation with an RGB-D camera
Software of the solution submitted by the team Visual at the 2022 ICASSP CORSMAL challenge.
[code] [paper]
Because It's Tactile Filling mass estimation using multi-modal observations of human-robot handovers
Software of the solution submitted by the Because It's Tactile team at the 2020 CORSMAL challenge.
[code] [paper] [video] [slides]
HVLR Audio-Visual Hybrid Approach for Filling Mass Estimation
Software of the solution submitted by the HVRL team at the 2020 CORSMAL challenge.
[code] [paper] [video] [slides]
NTNU-ERC NTNU-ERC solution for filling mass estimation
Software of the solution submitted by the NTNU-ERC team at the 2020 CORSMAL challenge.
[code] [slides]