Deep Learning for Clothing Style Recognition Using YOLOv5.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Deep learning advancements necessitate powerful hardware, posing challenges for resource-constrained users.
• Lightweight algorithms and accessible development environments are crucial for broader deep learning adoption.
• Cross-domain applications of deep learning are gaining significant traction in both academia and industry.

Purpose of the Study:

• To evaluate the YOLOv5s algorithm, a lightweight deep learning model, for object detection tasks.
• To investigate the performance of YOLOv5s in recognizing diverse clothing styles.
• To demonstrate the utility of Google Colab as an open-source environment for training and testing deep learning models.

Main Methods:

• Utilized the YOLOv5s algorithm, a one-stage object detection model.
• Trained and tested the model using Google Colab, an accessible cloud-based platform.
• Collected and categorized image data of fashion clothing from dedicated datasets and web crawling into five styles: plaid, plain, block, horizontal, and vertical.

Main Results:

• YOLOv5s achieved high recognition accuracy and fast detection speeds for clothing styles.
• Performance metrics included average precision, mean average precision, recall, F1-score, model size, and frames per second.
• Experimental outcomes indicated YOLOv5s outperformed other learning algorithms in accuracy and speed.

Conclusions:

• YOLOv5s offers an efficient and effective solution for object detection, particularly in scenarios with limited computational resources.
• The study highlights the advantages of one-stage object detection algorithms for practical applications like fashion analysis.
• Google Colab provides a viable and supportive open-source environment for developing and comparing deep learning models.