Detecting Pesticide Contamination in Fruits and Vegetables with Deep Learning Models

Pesticides are frequently mixed with various perishable fruits and vegetables in the market. Dishonest vendors often use these harmful chemicals to artificially preserve fruits for extended periods. Such adulterants are pretty hard to detect they do not change the outer appearance of the fruit. Therefore, a more sensitive and robust detection technique needs to be invented to handle such an issue efficiently. Therefore, in order to overcome this challenge, we propose a comprehensive deep-based detecting learning framework for fruits. For this experiment, different varieties of apples were tested in both contaminated states. The experiment was tested fresh and chemically toxic compounds present in fruit evaluation on pre-trained deep models and introduced a novel model for the purpose, which was ChemNet architecture-based. Firstly, the proposed ChemNet-based model's effectiveness was to be evaluated against existing deep learning models. The comparison was done in terms of accuracy of predicting a substance as a toxic one. The precision was found to be extremely low according to the outputs for GoogleNet and VGG-16, whereas YOLOv8 achieved a preciser 89.45. Surprisingly, the ChemNet-based architecture presented obtained better results than all other architectures with an accuracy rate of 97. 36 % in. These ChemN et are a resilient andsolution for safety in the market.