The goal of fruit and vegetable image recognition is to identify accurately these items in the photographs. The traditional applications utilize Convolutional Neural Networks (CNNs) with the MobileNetV2 architecture to efficiently recognize various fruits and vegetables. MobileNetV2’s efficient design enables deployment on computationally constrained Mobile and embedded devices. The model, trained on an extensive fruit and vegetable image dataset, uses a CNN with MobileNetV2 and employs advanced preprocessing and data augmentation techniques such as rotation and scaling to improve generalization. MobileNetV2 was preferred over MobileNet in this study because it incorporates inverted residual blocks with bottleneck features, supports input sizes greater than 32 \(\times \) 32, and has a lower parameter count. The experimental results show that MobileNetV2 offers improved accuracy and computational efficiency in fruit and vegetable recognition compared with traditional CNN models.