Download core samples from [Google Drive](https://drive.google.com/drive/folders/1i4sPhfbvCzV6b0r5iiUEnGVXmX1XkwMq), and extract `samples.zip` (sha256 - c1a27a0716b5633792afca7c1a032dcc9c15c8f7153a03e8d5206e1d86379896) in the project root under `samples` directory. Raw detection sample was large, thus it has been separated into `cctv.h264` file (sha256 - 9fd17d015924538c140f9ee478bdbfc5233f6c948ff134d2f31415e795da9b66). Place it in `samples/detection` directory. Locations are important, because predefined configs reference paths relative to current working directory.

Load predefined configuration files from the paths given (by clicking on the large detection title button).

- Scenario from Results - `samples/scenario_detect.json` (requires raw sample).

- Basic motion - `samples/basic_detect.json`.

- Live TCP stream (webcam) - `samples/tcp_detect.json`.

For V4L webcam through TCP, first run the following command in another terminal:

Adjust `/dev/video1` to correct V4L device. The command needs to be re-run after each connection instance.

Load predefined configuration files from the paths given (by clicking on the large tracking title button).

- Synthetic videos, all estimators, ground truth comparisons - `samples/synthetic_all_gt.json`.

- Synthetic videos, single estimator, ground truth - `samples/synthetic_almeida_gt.json`.

- Real-world videos - `samples/real_world.json`.

- Real-world videos (cropped, slowmo) - `samples/real_world_crop.json`.

- Live TCP stream (webcam) - `samples/tcp_track.json`.

The preloaded samples can be replaced with different ones. See `samples/synthetic` and `samples/real_world` directories. Cropped config is needed to account for 1.2x sensor crop applied on slow motion video.

For TCP configuration, refer to the previous subsection for setup steps. In addition, camera's horizontal and vertical field of view must be set correctly for tracking to be accurate.