Softwares and Datasets

The GAMMA research group has a variety of software and models available to download.

• Algebraic and Geometric Software
• Audio Simulation and Speech Processing
• Audio-Visual Processing
• Autonomous Driving
• Cache-coherent Layout Computation
• Cloth Simulation
• Collision Detection and Proximity Queries
• Crowd Simulation
• Emotion Classification
• GPGPU and Many-core Computing
• Graphics Programming Libraries
• High-DOF Grasping
• Models
• Motion Planning for Single Robot or Multiple Robots/Agents
• Physically-based Modeling, Simulation, and Animation
• Texture Encoding
• Indoor Navigation: Perception and Planning
• Outdoor Navigation: Perception and Planning
• Scene Understanding

Algebraic and Geometric Software

• BOOLE: Constructive solid geometry (CSG) models into boundary representation (B-rep) form
• Fast back-face culling using normal masks
• IK: Inverse kinematics code for serial manipulators
• MAPC: Efficient and exact manipulation of algebraic points and curves
• MARS: Maple/MATLAB/C resultant-based solver
• Polygon Triangulation: Fast polygon triangulation based on Seidel’s algorithm

Autonomous Driving

• Behavior-Driven Traffic Simulator: A traffic simulation software based on SUMO that incorporates driver behavior
• TrackNPred: Implementations of state-of-the-art tracking and trajectory prediction methods and tools to benchmark and evaluate them on real-world dense traffic datasets
• TRAF: Dataset of dense and heterogeneous traffic videos, consisting of multiple camera viewpoints (front-facing/top-view), camera motions (moving/static), times of day, and agents (car, bus, truck, rickshaw, pedestrian, scooter, motorcycle)
• METEOR: A new and complex traffic dataset for unstructured scenarios in India.

Cache-coherent Layout Computation

• OpenCCL: Cache-coherent layouts of meshes and graphs

Cloth Simulation

• I-Cloth: API for high-quality cloth simulation with CUDA

Collision Detection and Proximity Queries

• DEEP: Dual-space expansion for estimating penetration depth
• DEFORMCD: Collision detection for deforming objects
• FCL: A fast collision library and its integration with ROS for use with robotic systems
• MCCD: A library for multi-core collision detection
• gProximity: Hierarchical GPU-based operations for collision and distance queries
• PIVOT2D: Proximity information from Voronoi techniques
• PQP: Fast proximity queries with swept sphere volumes
• RAPID: Robust and accurate polygon interference detection
• SELF-CCD: Continuous collision detection for deforming objects
• BSC: Reliable continuous collision detection
• SWIFT: Speedy walking via improved feature testing
• SWIFT++: Speedy walking via improved feature testing for non-convex objects
• V-COLLIDE: Accelerated collision detection for VRML

Crowd Simulation

• India-Walk: Large Crowds Dataset captured in India
• Menge: Crowd Simulation Framework

Emotion Classification

• DeceptiveWalk: Dataset of 16-joint gaits representing deceptive and non-deceptive walking
• GroupWalk: Dataset of walking crowd videos collected in eight real-world locations
• TAEW: A semi-supervised method to detect one or more perceived emotions in gaits, between happy, sad, angry, and neutral
• STEP: A method to both generate synthetic emotive gaits and classify emotive gaits into one of four emotions, happy, sad, angry, and neutral
• Emotion-Gait: Dataset of 21-joint gaits annotated with four perceived emotions, happy, sad, angry, and neutral
• EmotionWalk: Dataset of 16-joint gaits annotated with four perceived emotions, happy, sad, angry and neutral

GPGPU and Many-core Computing

• GPUFFTW: High performance power-of-two fast Fourier transforms (FFT) using graphics processors
• GPUSORT: High performance sorting using graphics processors
• HAVOC3D: Fast distance field and Voronoi diagram computation using graphics hardware
• LU-GPU: Efficient algorithms for solving dense linear systems on graphics hardware

Graphics Programming Libraries

• GLVU: C++/OpenGL/GLUT-based viewer

High-DOF Grasping

• Dataset

Models

• Dynamic benchmarks
• Industrial: Industrial and CAD models (static scenes)

Motion Planning for Single Robot or Multiple Robots/Agents

• ITOMP Library: Realtime optimization based High-DOF planner for dynamic scenes
• HRVO Library: Independent navigation of multiple robots or virtual agents using hybrid reciprocal velocity obstacles
• RVO2 Library: Reciprocal collision avoidance for real-time multi-agent simulation
• D-ORCA: Distributed Optimal Reciprocal Collision Avoidance

Physically-based Modeling, Simulation, and Animation

• LumosQuad: Fast animation of lightning using an adaptive mesh

Audio Simulation and Speech Processing

• GSound: Fast sound propagation and rendering for games
• pygsound: Python package for impulse response generation based on GSound
• FAST-RIR: Python package for neural-network-based fast diffuse room impulse response generator
• MESH2IR: Python package for mesh-based neural network (MESH2IR) to generate acoustic impulse responses (IRs) for indoor 3D scenes represented using a mesh
• TS-RIR: Python package to translate synthetic RIR to real RIR
• IR-GAN: WaveGAN extension for room impulse response augmentation
• 6-channel synthetic IR dataset: Dataset used in our paper Improving Reverberant Speech Training Using Diffuse Acoustic Simulation
• Low-frequency compensated synthetic IR dataset: Dataset used in our paper: Low-frequency Compensated Synthetic Impulse Responses for Improved Far-field Speech Recognition
• MultiVox: A Benchmark for Evaluating Voice Assistants for Multimodal Interactions
• MMAU: A Massive Multi-Task Audio Understanding and Reasoning Benchmark
• GAMA: A large audio-language model with advanced audio understanding and complex reasoning abilities
• Audio-Flamingo 2 (AF2): A 3B parameter audio-language model with state-of-the-art audio understanding and long-form audio reasoning capabilities
• Audio-Flamingo 3 (AF3): A fully open-source audio-language model with unified speech/sound/music understanding, chain-of-thought reasoning, and long audio processing up to 10 minutes
• COMPA: A benchmark and training method for evaluating and improving compositional reasoning in audio-language models, including order understanding and attribute binding
• MMAU-Pro: A comprehensive benchmark with 5,305 expert-curated instances evaluating AI audio intelligence across 49 skills including speech, sound, music, and multi-audio reasoning

Audio-Visual Processing

• MeLFusion: A model for synthesizing music from combined textual descriptions and corresponding images
• Meerkat: An audio-visual language model with fine-grained spatial and temporal understanding of image and audio
• AURELIA: Actor-critic framework for test-time audio-visual reasoning with AVReasonBench evaluation benchmark
• AVTrustBench: Assessing and Enhancing Reliability and Robustness in Audio-Visual LLMs
• APoLLo: Unified Adapter and Prompt Learning for Vision Language Models

Texture Encoding

• GST: GPU-decodable Supercompressed Textures
• FasTC: Fast Texture Compression

Video Manipulation

• VideoSham: Video Manipulations Beyond Faces: A Dataset with Human-Machine Analysis

Indoor Navigation: Perception and Planning

Outdoor Navigation: Perception and Planning

• GND: GND: Global Navigation Dataset
• MTG: MTG: Mapless Trajectory Generator with Traversability Coverage for Outdoor Navigation
• DTG: DTG: Diffusion-based Trajectory Generation for Mapless Global Navigation
• AdaptiveON: Adaptiveon: Adaptive outdoor local navigation method for stable and reliable actions

Scene Understanding

• ET-Former: ET-Former: Efficient Triplane Deformable Attention for 3D Semantic Scene Completion From Monocular Camera
• RGN-D Indoor Place Recognition: Code for two papers: PoCo and CSCPR