My Role

Background : This project focuses on detecting and localizing alpacas in images using YOLOv8, a cutting-edge object detection algorithm. Leveraging a compact dataset and modest computational resources, I successfully trained a model for alpaca detection, achieving promising results despite the constraints.

Dataset : The dataset for this project comprises a small number of alpaca images, divided into training and validation sets. Given the limited size, I employed data augmentation techniques—such as flipping, rotation, and scaling—to enhance the variety and robustness of the training samples.

Methodology :

Data Preprocessing: Images were resized to a consistent input size compatible with the YOLOv8 architecture. Annotations were adjusted accordingly to match the resized images.
Model Configuration: Configured YOLOv8 with appropriate model architecture, hyperparameters, and anchor boxes tailored for alpaca detection.
Training: Conducted training over 50 epochs, monitoring key metrics such as loss and mean Average Precision (mAP). The limited computational resources were a factor in defining the number of epochs.

Results : The YOLOv8 model demonstrated encouraging performance in detecting alpacas within images, showcasing the potential of object detection even with a small dataset. While the detection accuracy and precision were not as high as those of models trained on larger datasets, the results affirm the feasibility of training effective models with constrained resources.