Comparison of Deep Learning-Based Object Classification Methods for Detecting Tomato Ripeness

Examination of the technological development in agriculture reveals that not many applications use cameras to detect tomato ripeness; therefore, tomato maturity is still determined manually. Currently, technological advances and developments are occurring rapidly, and are, therefore, also inseparable from the agricultural sector. Object detection can help determining tomato ripeness. In this research, faster region-based convolutional neural network (Faster R-CNN), single shot multibox detector (SSD), and you only look once (YOLO) models were tested to recognize or detect tomato ripeness using input images. The model training process required 5 hours and produced a total loss value <0.5, and as the total loss became smaller, the predicted results improved. Tests were conducted on a training dataset, and average accuracy values of 99.55%, 89.3%, and 94.6% were achieved using the Faster R-CNN, SSD, and YOLO models, respectively.

Automated summary

Examination of the technological development in agriculture suggests that there is a lack of applications using cameras to detect tomato ripeness, resulting in the need for manual determination. This research explores the use of faster region-based convolutional neural network, single shot multibox detector, and you only look once models to recognize or detect tomato ripeness, achieving high accuracy rates.