# SDC Assignment 5 - Argoverse Tracking
###### tags: `Self-Driving Cars`, `ROS`

In this assignment, you will need to run the baseline tracking module provided by Argo and visualize the tracking result using ROS.

## Goal
1. Understand file structure of an **open datasets**
2. Learn to utilize **baseline tracker**
3. Learn to **visualize** the result of tracking

## Data

### [Download Detection](https://s3.amazonaws.com/argoai-argoverse/detections_v1.1b.zip)
### [Download Pose-Only Dataset](https://drive.google.com/file/d/1-rLvNYIIMGuutnHfnVENUfr3KrR6zUVL/view?usp=sharing)
### [Download Full Data](https://www.argoverse.org/data.html#argoverse-11)

In this tracking assignment, we use only the **detection** result provided by Argo. In the future competition, you can use more information from the dataset if needed.

## Code

The baseline code is provided by Argo, which can be found [here](https://github.com/johnwlambert/argoverse_cbgs_kf_tracker).

## Visualization

We provide a visualziation tool for you. Dowload the package [here](https://drive.google.com/file/d/1ArJpFNwtsepa2_CYwWEUjXoB9UIqbisS/view?usp=sharing). You are required to run baseline code and record a video showing the IDs of each instance detected.

### Example:
[](https://youtu.be/YuEWxGrlNaQ)

## Bonus

If you can **draw the past trajectories** of tracked objects, you will get extra scores.
The trajectories of different objects should be easily identified by human eyes, namely, they should have **different colors/shapes or other properties** to distinguish one from another.

## Submission

Please Hand-in a video file showing your result, including:
1. LiDAR point cloud
2. Detection bounding box
3. Instance ID on each bounding box
4. (bonus) object trajectories

Naming Rule: `<student_id>_hw5.mp4`
Deadline: 12/8 22:00
**Note: The due time is extended to 12/8**