CNN Based Image Classification of Malicious UAVs

Round 1

Reviewer 1 Report

In this paper, the authors classify UAVs using different transfer learning architectures. Although the topic is interesting, author should do the following issues:

1. Introduction chapter is still poor. You should highlight what are the contributions in this paper. What kind of problem do you want to figure out in this draft? It is totally missing here. 

2. We can see accuracy is 85% and still the results are too poor. Why the results are not up to the mark? 

3. You should compare with other state-of-art models with your dataset as well as with other benchmark datasets

3. Nowadays, people are using a domain-specific dataset for implementing transfer learning, see the sample papers:

https://www.sciencedirect.com/science/article/abs/pii/S0263224119301939

https://link.springer.com/article/10.1007/s11227-022-04830-8

4. The dataset is too small as there are 127 images. Your results are not trustable due to limited samples. You should do the experiments with a large benchmark dataset. In addition, you also can apply data augmentation to increase the volume of the dataset.

You also should change the title.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The Paper discuss 18 the collection of images for three popular UAVs at different elevations and different distances from 19 the observer, and using different camera zoom levels. More detailed comments are given as follows:

1-     Why You select the pretrained ResNet-20, AlexNet, MobileNet and VGG16, what about other types such as, GOOGLENET, VGG19, ..etc.

2-    Give the details of the Accuracy measures.

3-    In A. DATASETS section, what about other DB.

4-    In training phase, dataset partition is randomly or not? I suggest u to used K-Fold Cross Validation. And how many K-Fold used?

5-     In experiments results, the evaluation (train-test round) must repeated for N round. I suggest to repeat for many rounds to ensure that the bias was minimized.

6-    The author(s) does not mention anything about the time for training phase which is important factor. I suggest to add some paragraphs about time analysis.

7-  Discuss the limitations of the proposed method.

8-       In conclusion section, put the future.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

In this paper, the authors applied transfer learning methods for UAV detections. In this paper, the author should propose a model and should compare the model with other state-of-art models. Should include a clear block diagram of the proposed model.

Author Response

Please see the attachment.

Author Response File: Author Response.docx