[Archived - 2021] Custom Object Detection: Target Class Recognition (YOLOv4/v3)
An archived technical test from 2021. This project demonstrates the complete end-to-end pipeline for training a custom object detection model to recognize a specific target object.
The focus was on creating a tailored dataset from scratch, annotating the unique, text-based object ("Agada"), and fine-tuning a YOLO-based architecture. The model successfully performs custom class detection with high confidence in a real-world scenario.
Note: Inference optimized for real-time performance on desktop hardware using CUDA. Face blur applied for privacy.
Technical Stack:
Python
Custom Dataset Collection & Annotation
Transfer Learning & Fine-tuning (YOLO)
CUDA Acceleration
#ComputerVision #YOLO #CustomObjectDetection #AITraining #MachineLearning #DeepLearning #Python #ImageAnnotation #CustomClass #AIEngineering #Archived