Real-Time Dress Code Detection using MobileNetV2 Transfer Learning on NVIDIA Jetson Nano
Laxmi Kantham Durgam, Ravi Kumar Jatoth · 2023
The proposed research intends to develop a model for automatically identifying the dress code in companies and educational institutions where appropriate apparel needs to be frequently maintained. The concept focuses particularly on the dress code for school children. Convolutional Neural Networks (CNN), MobileNetV2, and an object detection algorithm with a high level of accuracy are examples of deep learning models. In comparison to manual checks, using an automated system for dress code detection has various advantages. It reduces the number of mistakes and faults that could happen during manual inspections while also saving time. Organizations and institutions can make sure that their staff constantly complies with the dress code rules thanks to the automated system. The Dress code detection model is designed, trained, and implemented using the Edge impulse API. The proposed Transfer learning MobileNetV2 and CNN TINY ML models were implemented on an NVIDIA Jetson Nano edge device for real-time Dress code classification. F1 Score, precision, recall, latency, and storage space are common evaluation metrics for object detection models. The goal of this project is to create a model for automatically identifying dress rules in businesses and educational institutions that must maintain suitable attire on a regular basis. The proposed model offers a practical approach to tracking workplace and educational dress code observance. A disciplined environment is created and the professionalism of the individuals inside the organization is reflected by the accuracy of the MobilenetV2 and CNN models in detecting business formal clothes. The Proposed transfer learning MobileNetV2 model performs better results compared to all other CNN models in accurately detecting business and school children's formal attire.