Efficient Deep Learning Methods for Identification of Defective Casting Products

EasyChair Preprint 7904

Abstract

Quality inspection has become crucial in any large-scale manufacturing industry recently. In order to reduce human error, it has become imperative to use efficient and low computational AI algorithms to identify such defective products. In this paper, we compared and contrasted various transfer learning architectures and custom-built architectures for detecting defective casting products. Our results show that custom architectures transfer learning architectures in evaluation metrics. Moreover, custom models perform 6 to 9 times faster than lightweight models such as MobileNetV2 and NASNet. The number of training parameters and the model size of the custom architectures is significantly lower (~386 times & ~119 times respectively) than the best performing models such as MobileNetV2 and NASNet. Augmentation experimentations have also been carried out on the custom architectures to make the models more robust and generalizable. Our work sheds light on the efficiency of these custom-built architectures for deployment on Edge and IoT devices and that transfer learning models may not always be ideal. Instead, they should be specific to the kind of dataset and the classification problem at hand.

Keyphrases: Convolutional Neural Network, Efficient Transfer Learning, MobileNet, NASNet, ResNet, deep learning, defect detection, defective casting product, inference time, neural network

@booklet{EasyChair:7904,
  author    = {Bharath Kumar Bolla and Mohan Kingam and Sabeesh Ethiraj},
  title     = {Efficient Deep Learning Methods for Identification of Defective Casting Products},
  howpublished = {EasyChair Preprint 7904},
  year      = {EasyChair, 2022}}