Examples
Runnable demos live under examples/
in the repository, with a Python and a Rust entry per tracker.
| Tracker | Python | Rust |
|---|---|---|
| ByteTrack | byte_track_demo.py | byte_track_demo.rs |
| DeepSORT | deepsort_demo.py | deepsort_simple.rs, deepsort_ort.rs |
| OC-SORT | ocsort_demo.py | - |
| Deep OC-SORT | deep_ocsort_demo.py | - |
| BoT-SORT | botsort_demo.py | det_ind_demo.rs |
| SORT | sort_yolo_demo.py, sort_rtdetr_demo.py | - |
| TrackTrack | tracktrack_demo.py | tracktrack_demo.rs |
# Python
python examples/python/byte_track_demo.py
# Rust
cargo run --example byte_track_demo
cargo run --example deepsort_simple --features reid-model
cargo run --example deepsort_ort --features advanced_examples
cargo run --example det_ind_demo
cargo run --example tracktrack_demo
The Python demos use the usual detector stacks (ultralytics, transformers + torch,
torch + torchvision); install what a given demo imports. The deepsort_simple Rust demo
needs the reid-model feature and the deepsort_ort demo needs the advanced_examples
feature (ONNX Runtime + OpenCV).
Ultralytics
Python: trackforge.track_ultralytics(model, source, tracker) runs an Ultralytics model over a video,
webcam, stream or image folder and yields (result, tracks) per frame, and
trackforge.from_ultralytics(result) converts a single result for your own loop. See
ultralytics_track.py.
for result, tracks in trackforge.track_ultralytics(YOLO("yolo11n.pt"), "people.mp4", trackforge.BYTETRACK(), classes=[0]):
print(len(tracks))
Rust: trackforge::utils::detections_from_xyxy turns the [x1, y1, x2, y2, conf, cls] box data of
ultralytics-inference boxes into tracker input. The
examples/ultralytics
crate tracks a video or image folder with any motion tracker. It is a separate crate because
ultralytics-inference is AGPL-3.0.
cd examples/ultralytics
cargo run --release -- "frames/*.jpg" --tracker bytetrack --classes 0
cargo run --release --features video -- people.mp4 --classes 0 --save tracked.mp4