Description: Develop a model to help identify invasive species such as the crown of thorns starfish using imagery.

The team performed the following activities:

Set up Kaggle Notebook for Training and Inference​

Model Selection-Team selected YOLOv5: A Family of Object Detection Models

Used Provided Code to Perform Multiple Runs with Various Parameters

Used Various Hyperparameters for Tuning (e.g. rotation, translation, flipping)

Tracked Parameters and Scores

Use Model Outputs to Feed into an Inference Notebook

Submitted the Inference Notebook to be Graded Off a Private Test Dataset

Results: Used approximately 23,000 images in the training set to train a model to analyze approximately 13,000 images in the test set resulting in a 64% identification rate. This detection rate would theoretically enable a more efficient and cost-effective ability to locate the crown of thorns starfish, and to help researchers take better-informed actions to protect the reef.​