Vision-based Object Classification using Deep Learning for Inventory Tracking in Automated Warehouse Environment

To achieve automatic inventory management in warehouses, it is necessary to identify items. Barcodes and RFID tags are traditional approaches to solve this problem but both of them suffer from limitations. This research paper presents a vision-based method using a deep convolutional neural network to classify different items stored in a warehouse for the purpose of inventory management. The proposed method uses residual learning and employs ResNet-50 network architecture. It achieves a high accuracy of 98.94% on the dataset created by the authors consisting of 1450+ images of machine parts, by utilizing data augmentation and transfer learning. It runs at 4 frames per second (FPS), making it suitable for other real-time applications as well.