Driving Video with Object Tracking
A set of driving videos with multi-object tracking annotations, ideal for training computer vision models applied to mobility.
Description
This dataset offers 1002 driving video sequences accompanied by annotations for tracking multiple objects. Each video includes CSV files listing the positions of monitored objects (cars, pedestrians, etc.) frame by frame. It is designed to facilitate research in the fields of embedded computer vision, mobile robotics, and AI for autonomous vehicles.
What is this dataset for?
- Develop algorithms for tracking objects in real road environments
- Training video detection models for autonomous vehicles or embedded cameras
- Testing real-time detection algorithms in urban scenarios
Can it be enriched or improved?
Yes. It is possible to add additional classes of objects, to annotate weather conditions or road types, or to use it to train trajectory prediction models. Combined use with frameworks like YOLO, DeepSort or MMTracking is also possible.
🔎 In summary
🧠 Recommended for
- Computer vision engineers
- University projects
- R&D in smart mobility
🔧 Compatible tools
- OpenCV
- YoloV8
- MMTracking
- DeepSort
- Detectron2
💡 Tip
Start by viewing the annotations with a tool like CVAT or FiftyOne to validate the quality of the follow-up.
Frequently Asked Questions
Does this dataset only contain cars?
No, it includes several types of mobile objects such as cars, pedestrians, or cyclists, depending on the scenes recorded.
Are the annotations synchronized frame by frame?
Yes, each associated CSV file specifies the coordinates of the objects for each frame, facilitating tracking tasks.
Can this dataset be used for real-time detection?
Yes, videos and annotations are compatible with real-time frameworks like YOLO or DeepSort, ideal for this type of use.
