Deep learning YOLOv5s Model for Tiny Pests Detection

The whiteflies are one of the most destructive tiny insects found in high-yielding crops, and because of their wide range of spread, they can seriously harm agriculture. Because these pest species cause insufficient agricultural production on a large scale, it is challenging and time-consuming for humans to observe these insects in a larger field. Recently, deep learning has been applied to agriculture to detect plant pests. This study proposes using the YOLOv5s convolutional neural network algorithm to identify whiteflies on plant leaves and add this data to the whitefly dataset repository. A comparative analysis was conducted to evaluate the effectiveness of the model when subjected to the pre-processing technique of tilling, as well as data enhanced by random rotation and random cropping. The speed and AP score (AP 50 and AP 50–95) were used in the comparison process. According to the experimental results, the YOLOv5s algorithm is effective in augmenting data through random crop. The AP 50 of YOLOv5s can therefore be raised from 94 % to 98%, which is adequate to meet the requirements for recognising and detecting whiteflies even though they look tiny and very similar to white spots caused by water droplets, dust on leaves, or their light reflection.