Fruit Classification using Colorized Depth Images

Fruit classification is a computer vision task that aims to classify fruit classes correctly, given an image. Nearly all studies have used RGB color images as inputs, few costly hyperspectral images, and classical ML-based colorized depth images. Depth apparent benefits such invariance lighting, less storage requirement, better foreground-background separation, more pronounced curvature details object edge discontinuities. However, the use of in CNN-based remains unexplored. The purpose this study investigate with four CNN models, namely, AlexNet, GoogleNet, ResNet101, VGG16, compare their performance computational efficiency, well impact transfer learning. apple, orange, mango, banana rambutan (Nephelium Lappaceum) were manually collected using sensor sub-millimeter accuracy subjected jet, uniform, inverse colorization produce three sets dataset. Results show can be train models for ResNet101 achieving best 96%on It achieved 100% after GoogleNet showed most significant improvement learning on uniform dataset, at 12.27%. also exhibited lowest training inference times. results potential similar tasks.