Leaf miner pests seriously threaten the productivity, profitability, and sustainability of soil-less tomato cultivation systems. Early and accurate identification of leaf miner infestation is crucial for timely pest control measures. 

A study by researchers at the National Institute of Technology Hamipur and ICAR—Central Institute of Agricultural Engineering, Bhopal, presents an efficient approach using attention-based convolutional neural networks to identify this pest infestation in a timely manner. Different hyperparameters were tuned to get the optimal model performance. The custom model was trained using an image dataset collected from tomatoes grown in a hydroponic system. The proposed attention-based CNN model achieved an overall accuracy of 97.87%, 97.10% precision, 98.53% recall, and environment.

Analyzing the results

The experimental results show that the proposed attention-based CNN model achieved an overall accuracy of 97.87%, 97.10% precision, 98.53% recall, and VGG19. It outperformed state-of-the-art CNN models due to its improved feature extraction capability. The 97.81% F1-score. Additionally, the model performance was compared with other pre-trained models viz., AlexNet, VGG16, and the efficiency of the model underlines its potential to be deployed as part of automated pest monitoring systems in hydroponic VGG19 and was found to outperform these state-of-the-art CNN models due to its improved feature extraction capability. This work contributes to the development of computer vision and deep learning-based solutions for the precision efficiency of the model, which underlines its potential to be deployed as part of automated pest monitoring systems in hydroponic environments.

To demonstrate the model's efficiency, several experiments were conducted by varying different parameters related to the attention-based CNN model. The optimal model is found using extensive hyperparameter tuning in the first stage. For the seamless training of the model, the dataset was exported into Google Drive and mounted to the Google Colab platform. The developed attention-based DCNN model possesses 103,694,658 total trainable parameters. The model size was observed to be 239 MB. All the models were trained for 200 epochs, and two different approaches were used to prevent overfitting, viz., data augmentation and early stopping. During the training process, all the models converged during early epochs without overfitting. 

Targeted and cost-effective

The objective behind developing an automated approach for the automatic identification of leaf miners is due to the increased destruction power of these pests in hydroponic environments. Early identification of leaf miners can allow for more targeted and cost-effective pest management with minimal use of chemicals, thus reducing the overall environmental impact. However, this study is limited to binary classification, i.e., classification into healthy and leaf miner classes.

Full Report