Custom Dog/Cat Object Detection Dataset (Project)
I developed a custom multi-class object detection dataset for animal class identification, specifically for dog and cat detection models. My responsibilities included dataset sourcing, annotation, export in YOLO format, and ground-truth validation against model predictions. These tasks improved mean average precision (mAP) and overall annotation consistency. • Established full class mapping and dataset splits for training/validation/test. • Validated and corrected inconsistent annotations to enhance performance. • Utilized Label Studio for structured annotation and export. • Supported model evaluation through detailed comparison with ground truth.