The EE 443 2024 Challenge: Single Camera Multi-Object Tracking aims to enhance the performance of object detection and tracking algorithms in single-camera environments. Participants will focus on improving detection models, ReID (Re-identification) models, and Multi-Object Tracking (MOT) algorithms to achieve superior object tracking accuracy.
- Release of the Challenge & Data: May 3rd, 2024
- Team Registration Due: May 8th | 11:59:59 pm
- Release of the Baseline Code: May 9th | 11:59:59 pm
- Final Submission (Results) Due: June 3rd | 11:59:59 pm
- Final Presentation (in-person): June 4th & June 6th
- Github & Final Report Due: June 7th | 11:59:59 pm
-
Install ultralytics (follow the Quickstart - Ultralytics)
-
Download the
data.zipfrom GDrive link provided in the Ed Discussion
Your folder structure should look like this:
├── data
│ ├── test
│ ├── train
│ └── val
├── detection
│ ├── 1_prepare_data_in_ultralytics_format.py
│ ├── 2_train_ultralytics.py
│ ├── 3_inference_ultralytics.py
│ └── ee443.yaml
- Prepare the dataset into ultralytics format (remember to modified the path in the script)
python3 detection/1_prepare_data_in_ultralytics_format.py
After the script, your ultralytics_data folder should looke like this:
├── data
├── detection
├── ultralytics_data
│ ├── train
│ │ ├── images
│ │ └── labels
│ └── val
│ ├── images
│ └── labels
- Train the model using ultralytics formatted data (remember to modified the path in the script and config file
ee443.yaml)
python3 detection/2_train_ultralytics.py
You model will be saved to runs/detect/ with an unique naming.
- Inference the model using the testing data (remember to modified the path in the script)
python3 detection/3_inference_ultralytics.py