GTAVDataCollection is a mod to extract synthetic data from Grand Theft Auto V. The data includes photo-realistic computer images and annotations that can be used for the training of machine learning algorithms.
English | 简体中文
- ScriptHookV
- ScriptHookVDotNet
- Scripted Camera Tool
- Visual Studio 2019
- .NET Framework ≥ 4.8
- Others managed by Nuget
- Download and install ScriptHookV、ScriptHookVDotNet and Scripted Camera Tool.
- Download pre-built binaries and copy them to
Grand Theft Auto V/scripts/
. - Start the game.
- Press T to set the camera.
- Press Y to start or stop extracting data. The data should be created in
Grand Theft Auto V/data/
.
A large-scale remote sensing image dataset for vehicle object detection called GTA5-Vehicle has been built based on GTAVDataCollection. Please also note that the data is for research and educational use only.
The dataset includes 15 vehicle object classes: Compacts, Sedans, SUVs, Coupes, Muscle, SportsClassics, Sports, Super, OffRoad, Industrial, Utility, Vans, Service, Emergency, and Commercial. Here are some example images of these objects:
Compacts | |||||
Sedans | |||||
SUVs | |||||
Coupes | |||||
Muscle | |||||
SportsClassics | |||||
Sports | |||||
Super | |||||
OffRoad | |||||
Industrial | |||||
Utility | |||||
Vans | |||||
Service | |||||
Emergency | |||||
Commercial |
Each image has a corresponding label file. Here’s an example of a label structure: The first line contains the image’s dimensions. The second line provides camera pose information, including position and orientation. From the third line onwards, each line represents a target with its index, bounding box coordinates, vehicle category, size, and quality
2048,1152
X:-420.2515 Y:-2069.219 Z:144.7112,X:-89.97202 Y:-70.18397 Z:0
0,1120,255,1149,331,Sedans,small-vehicle,High
1,1100,732,1126,796,Sedans,small-vehicle,High
2,3,701,92,801,Commercial,large-vehicle,High
...
If you make use of the GTAVDataCollection or GTA5-Vehicle dataset, please cite our following paper