Crafting Diverse Synthetic Worlds for AI Perception.
GenVista is an open-source, Unity-based framework for generating diverse, richly annotated synthetic datasets. It aims to help ML, robotics, and computer vision developers create customizable 3D scenes and simulate sensor outputs to accelerate AI model development. Our mission is to provide a versatile, community-driven tool for a wide array of perception tasks. It has just started, but it will be upgraded from time to time.
- Versatile 3D Environments: Import real-life 3D scenarious, using CESIUM for Unity.
- Customizable Sensors: Simulate cameras (RGB, depth, segmentation) with extensibility for more. (TBD)
- Rich Annotations: Auto-generate 2D/3D bounding boxes, semantic/instance masks, and depth maps. (TBD)
- Domain Randomization: Tools to vary textures, lighting, and object placements. (TBD)
- Unity Powered: Leverages Unity for high-fidelity rendering and physics.
- Scripting & Extensibility: Full programmatic control and modular design for community contributions.
- Detailed Atmosphere: Thanks to Tenkoku Dynamic Sky, it is simple to edit the weather conditions.
- Prerequisites: Unity Hub, Unity Editor (2022.3 LTS+), Git.
- Clone:
git clone https://github.com/EmanuelCostaS/GenVista.git - Open in Unity: Use Unity Hub to open the cloned project.
(To be added.)
- Generating datasets to train object detectors for autonomous systems (vehicles, drones).
- Developing perception algorithms for robotics in varied environments.
- Generating large-scale datasets for computer vision research (segmentation, depth estimation).
- Prototyping 3D assets and environments for simulations or VR/AR.
- Supplementing real-world data to improve model generalization.
We welcome community contributions! Please see CONTRIBUTING.md (to be added) for guidelines on reporting bugs, suggesting features, and submitting pull requests.
This project utilizes several key assets to achieve its functionality and visual fidelity. We are grateful to the creators for their valuable resources.
This project utilizes Cesium for Unity along with data streamed from Cesium ion to integrate real-world 3D geospatial capabilities. We thank the Cesium team for providing these powerful tools.
Important Usage Note for Cesium Platform: This project is developed for research and non-commercial purposes, utilizing a Cesium ion Community account. For any commercial use of Cesium ion data and services, a commercial Cesium ion subscription is required in accordance with Cesium's terms of service. Please refer to the Cesium ion pricing page for more details on account types and commercial licensing.
This project utilizes the Tenkoku Dynamic Sky for its dynamic sky, lighting, and weather effects. This is a free asset obtained from Unity Asset Store. It is used in accordance with its specified license: Unity Asset Store EULA for free assets. We thank Tanuki Digital for making this valuable resource freely available.
GenVista Engine is released under the Apache License 2.0,
See the LICENSE file for details. (TBD)
Thank you for your interest in GenVista! If you like it, please leave a star ⭐