Enabling Intelligence on the Edge: Leveraging Edge Impulse to Deploy Multiple Deep Learning Models on Edge Devices for Tomato Leaf Disease Detection

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article presents a solid exploration of deploying multiple deep learning models on edge devices for tomato leaf disease detection. It combines state-of-the-art technologies like Edge AI and TensorFlow, and shows promising results in terms of accuracy and efficiency. However, to enhance the practical applicability of the work, improvements in dataset diversity, model optimization for real-world conditions, and strategies for large-scale deployment could be considered.

1.The dataset primarily consists of images from a specific geographic location (Ghana), which may limit the model’s ability to generalize to other regions or variations in tomato diseases. Expanding the dataset to include more diverse sources and conditions (e.g., different climates or soil types) could enhance the model's robustness.

2.Models like MobileNet and EfficientNet showed signs of overfitting, with significant discrepancies between training and validation accuracies. This suggests that the model might perform poorly on unseen data. Implementing additional regularization techniques, such as dropout, or using cross-validation methods might help mitigate overfitting.

3.While the study focuses on deploying models on edge devices, some models (e.g., Inception) have large sizes and slow inference speeds, making them unsuitable for resource-limited environments. Further optimization techniques like pruning or knowledge distillation could be applied to reduce model size and improve efficiency without compromising accuracy.

4.Although the system is described as scalable and cost-effective, more details could be provided about the practical challenges of large-scale deployment, such as maintaining model accuracy in changing environmental conditions or integrating with existing agricultural infrastructure.

5. Some new reports related to this work may should be included in the references. Such as Sensors & Actuators: B. Chemical 326 (2021) 128991;Chemical Engineering Journal 452 (2023) 139504;Advanced Science 9 (2022) 2202505;Chemical Engineering Journal 498 (2024) 155355

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

. The introduction and literature review section has very limited information with two to three lines in a paragraph without much technical information. Rewriting with more content is required.

2.       Why it is required to go for TensorFlow Lite (TFLite) format and quantized to 8-bit integers apart from the speed. Why integers?

3.       There are different varieties of tomato are present in the world. Details related to the verity is missing. (Cherry tomatoe, Grape tomatoes, Roma tomatoes, Beefsteak tomatoes, Heirloom tomatoes etc). Mention the Varity name.

4.       The proposed methodology in Figure 2: Stages of the proposed method is not clear. Think of first-time reader.

5.       The quality of the Figure 3: (Samples of the dataset images.) is not good. A good quality images are needed.

6.       What is the tool has been used for the mobile application development?

7.       Author may consider the following reference if suites 1) Identification and classification of groundnut leaf disease using convolutional neural network- International Conference on Computational Intelligence in Data Science 2) Detection and classification of paddy leaf diseases using deep learning (cnn) International Conference on Computer, Communication, and Signal Processing.

8.       Image quality is poor. Author may take few sample images to differentiate diseases or good leaf.

9.       What about the processing time?

1.   Have you tested the real time data?

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have made good improvements to the manuscript, which I think will be accepted for publication.