Search

Over a decade, convolutional neural networks (CNNs) have been applied extensively on various tasks related to images. Given an input image, a CNN model will investigate the content and deduce the representation of this image using a model's structure built from hidden neurons. This representation analyzes data semantically, which helps to solve semantic issues, such as image retrieval. To verify the above viewpoint, this study addresses the problem of using features learned from a CNN model to perform image retrieval. To more emphasize the efficiency of learned features, we consider degraded images and their enhanced version as queries and search for similar ones in the gallery set. Data augmentation is also applied to increase the number of images in the gallery. The experiments are conducted on a multi-view dataset, smallNorb. Experimental results are reported both in quantity and quality.

In this paper, the defective product classification based on deep learning for a smart factory is introduced. The proposed system contains PLC (Programmable Logic Controller), Artificial Intelligence (AI) embedded board and cloud service. The AI embedded board is connected and communicated to receive and send commands to PLC via SPI (Serial Peripheral Interface) protocol. The pre-trained defective product classification model is uploaded, saved on a cloud server and downloaded to AI Embedded board for each particular product. The core technique of the system is the AI-based embedded board. Due to the limitation of label data, we use transfer learning method to retrain deep neural networks (DNN). We implement and compare the classification results on different deep neural network including ResNet, DenseNet, and GoogLeNet. We trained these networks by GPU server on casting product classification data. After that, the pre-trained models are optimized and applied on practical embedded board. The experimental results show that our system is able to classify defective products with high accuracy and fast speed.