YOLOv5: https://github.com/ultralytics/yolov5

Miniproject for TDT17 Fall 2021

• Kaggle: pip3 install kaggle (follow official instructions for authentication)
• Weights & biases: pip3 install wandb (follow official instructions for authentication)

1. Clone the repo: git clone https://github.com/ultralytics/yolov5.git.
2. Install the necessary dependencies: cd yolov5 & pip3 install -r requirements.txt.

Dataset can be found on kaggle: https://www.kaggle.com/bjosttveit/tdt17avcombined

Run ./data.sh to download the training data from kaggle.

Models can be found on weights and biases: https://wandb.ai/bjosttveit/TDT17

Run python3 models.py to download the custom models trained on the data.

To produce frames with bounding boxes overlaid, run: ./detect.sh [n|n6|s|s6|m|m6|l|l6|x|x6].

Run ./test.sh [n|n6|s|s6|m|m6|l|l6|x|x6] to evaluate a model on the test set.