Comparative Study of Image Classification Models Using Deep Learning: MobileNet, ResNet, and EfficientNet

Comparative Study of Image Classification Models Using Deep Learning: MobileNet, ResNet, and EfficientNet

Aditya Randive1, Shrayash Gawade2, Aditya Shinde3, Sarang Admane4, Prof. R.M. Wahul5

1Dept. of Computer Engineering, MES Wadia College Of Engineering, Pune, India

 2Dept. of Computer Engineering, MES Wadia College Of Engineering, Pune, India

 3Dept. of Computer Engineering, MES Wadia College Of Engineering, Pune, India

4Dept. of Computer Engineering, MES Wadia College Of Engineering, Pune, India

5Dept. of Computer Engineering, MES Wadia College Of Engineering, Pune, India

Abstract—Object detection is a crucial aspect of computer vision that enables effective interpretation of visual data. This paper presents a comparative study of three prominent deep learning models, EfficientNet, ResNet, and MobileNet, focusing on their performance in various object detection tasks. As the demand for accurate and efficient solutions grows, understanding the strengths and weaknesses of these models becomes essential. Our research evaluates models based on standard datasets such as COCO and Pascal VOC, analyzing key metrics like precision, recall, mean average precision (mAP), and inference time. The findings reveal important insights into the trade-offs between accuracy, speed, and computational efficiency. In addition, we explore the effects of transfer learning and hyperparameter tuning, demonstrating improvements in detection accuracy and training efficiency. This comparative study provides valuable information for researchers and practitioners in the field of object detection, helping to select the most effective models for various applications.

Index Terms—Object Detection, Deep Learning, EfficientNet, ResNet, MobileNet, Computer Vision, Machine Learning, Perfor- mance Evaluation, Transfer Learning, Mean Average Precision