A 3D Fluorescence Classification and Component Prediction Method Based on VGG Convolutional Neural Network and PARAFAC Analysis Method
, Shun Zhao 1, Xueqin Jiang 1, Yixuan Li 2, Jianbo Fei 3, Dinghua Ou 3, Qiang Tang 1, Zhiwei Lu 4, Tao Liu 1 and Jianguo Xia 3,*Round 1
Reviewer 1 Report
1) Are there any applications of deep convolutional neural networks or other deep learning structures in the literature for this classification (even for 2D)? if yes, also mention them in the review part of the introduction.
2) How were hyper-parameters tuned? For the training of a deep neural network, it is needed to use validation data set (in addition to the test ) for hyperparameter tuning. No validation set has been considered for the mentioned task.
3) Fig 8. The loss plot should show both train and validation losses. In this way, the plot is used for the monitoring the training process of the network to avoid overfitting. In Fig 8, only the train loss is depicted.
4) Table 3: Reaching the accuracy of 1 is suspicious. It might be because of using the same data for validation and testing. Divide the dataset to train, validation, and test and provide the evaluation based on the performance of the network on test data which is not used in the training of algorithms.
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The article presents the method for classification of 3D fluorescence images based on VGG network and parallel factor analysis (PARAFAC).
The layout of the paper includes:
- Introduction to the databases used for evaluation analysis as described in Table 1.
- Pre-processing of data before analysis.
- Proposed PARAFAC model for data analysis.
- Proposal to apply VGG network to evaluate data analysis
- Evaluate the operation of the proposed VGG network with the original database.
The reviewers found that the structure of the article was quite tight.
However, during the review of the article, the reviewer found that no 3D images were analyzed. The evaluation only focuses on the operation of the VGG model during the training process.
Therefore, the reviewer suggested that the author add another part only focusing on the analysis of 3D fluorescence images to highlight the content and results of the paper.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
the reviewer agrees with the author's response.
