📜 Description

A deep learning computer vision project for detecting and localizing balls in images using a custom CNN architecture. The system employs an approach with dual heads: binary classifier to detect ball presence and a regressor for precise bounding box localization.

⚙️ Features

• CNN Architecture – 4-layer convolutional backbone
• Binary Classification – Detects ball presence with a binary classifier
• Bounding Box Regression – Precise localization with normalized YOLO-format coordinates
• Data Augmentation – Image transformations for improved model generalization