Transfer‐learning‐based Raman spectra identification

Zhang, Rui; Xie, Huimin; Cai, Shuning; Hu, Yong‐Jie; Liu, Guokun; Hong, Wenjing; Tian, Zhong‐Qun

doi:10.1002/jrs.5750

Cited by 57 publications

(59 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Considering that CNN is a “data‐hungry” model, [ 12 ] it is necessary to perform data augmentation in each training process. After expanding the amount of data of training sets by 50 times, the accuracy of test sets of 1D‐CNN model significantly increases.…”

Section: Resultsmentioning

confidence: 99%

“…[ 10,11 ] Furthermore, it has been proved that CNN model trained with big data sets can achieve good identification results for Raman spectral data sets through transfer learning. [ 12 ] On this basis, the method proposed in this paper is very meaningful and makes it possible for CNN model to be applied to Raman spectral data sets with different spectral size, even if the data sets are collected by different spectrometers. This research further explores the requirement for calibration data and processing methods of 1D‐CNN model in transfer learning, which will really improve the application of 1D‐CNN in Raman spectroscopy.…”

Section: Discussionmentioning

confidence: 99%

“…In addition, Zhang's group tried applying transfer learning to RRUFF database and found that transfer learning technique has great potential in the effective identification of Raman spectra when the number of spectral data is limited. [ 12 ]…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A novel polynomial reconstruction algorithm‐based 1D convolutional neural network used for transfer learning in Raman spectroscopy application

Shang

Bao

Tang

et al. 2021

J Raman Spectroscopy

View full text Add to dashboard Cite

When a convolutional neural network (CNN) model was built, the size and resolution of its input data were fixed. However, Raman spectra collected by different Raman spectrometers usually had different length, intensity range, and wavenumber interval between two adjacent data points, which made the existing CNN model difficult to be applied to a new Raman spectral data set. Therefore, this paper proposed a polynomial reconstruction algorithm as pretreatment method to obtain reconstructed spectra that would be imported into CNN model with consistent length, intensity range, and wavenumber interval. To test the effectiveness of this method, a big data set with 2563 Raman spectra of 831 minerals and synthetic organic pigments samples was constructed from the RRUFF and SOP database to pretrain a one‐dimensional CNN (1D‐CNN) model. The pretraining results showed that polynomial reconstruction algorithm used as pretreatment method was better than SG smoothing combined spline interpolation algorithm. Then two data sets were collected by different Raman spectrometers for evaluating the transfer learning performance of the trained 1D‐CNN model. Both data sets contained 390 Raman spectra from the same 39 samples of inorganic salts, organic compounds, and amino acids. One was used as calibration data to retrain the 1D‐CNN model, while the other was used as test. Based on data augmentation and 75% calibration data for retraining, the transfer learning performances of 1D‐CNN model were clearly shown in the excellent identification accuracies of 99.58%, 99.32%, and 97.69% for training, validation, and test sets, respectively, which were better than those of K‐nearest neighbor classifier. This paper provides a significant way for the wide application of CNN model in Raman spectroscopy with much more advantages in simplicity and rapidity.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

A novel polynomial reconstruction algorithm‐based 1D convolutional neural network used for transfer learning in Raman spectroscopy application

Shang

Bao

Tang

et al. 2021

J Raman Spectroscopy

View full text Add to dashboard Cite

show abstract

“…Deep neural network models were performed in Keras, and its optimizer was Adam ( Zhang et al, 2019 ). The VAE encodes high-dimensional data (LC–MS profiles) into a low-dimensional latent space to select primary representations of the data.…”

Section: Methodsmentioning

confidence: 99%

Pseudotargeted Metabolomic Fingerprinting and Deep Learning for Identification and Visualization of Common Pathogens

et al. 2022

View full text Add to dashboard Cite

Matrix-assisted laser desorption/ionization time-of-flight mass (MALDI-TOF) spectrometry fingerprinting has reduced turnaround times, costs, and labor as conventional procedures in various laboratories. However, some species strains with high genetic correlation have not been directly distinguished using conventional standard procedures. Metabolomes can identify these strains by amplifying the minor differences because they are directly related to the phenotype. The pseudotargeted metabolomics method has the advantages of both non-targeted and targeted metabolomics. It can provide a new semi-quantitative fingerprinting with high coverage. We combined this pseudotargeted metabolomic fingerprinting with deep learning technology for the identification and visualization of the pathogen. A variational autoencoder framework was performed to identify and classify pathogenic bacteria and achieve their visualization, with prediction accuracy exceeding 99%. Therefore, this technology will be a powerful tool for rapidly and accurately identifying pathogens.

show abstract

“…It can migrate the knowledge of a pre-trained network based on the source dataset to the target dataset (Oquab et al, 2014 ). Considerable publications underline the benefits of pre-training deep networks on large datasets (Käding et al, 2016 ; Zhang et al, 2019 ). To conduct fine-tuning, the network is first trained on the source dataset, and then, pre-trained parameters are transferred to the target task and kept fixed, with only a few layers (commonly the last few layers) trained on the target dataset (Oquab et al, 2014 ).…”

Section: Methodsmentioning

confidence: 99%

Hyperspectral Imaging Combined With Deep Transfer Learning for Rice Disease Detection

Liu

et al. 2021

Front. Plant Sci.

View full text Add to dashboard Cite

Various rice diseases threaten the growth of rice. It is of great importance to achieve the rapid and accurate detection of rice diseases for precise disease prevention and control. Hyperspectral imaging (HSI) was performed to detect rice leaf diseases in four different varieties of rice. Considering that it costs much time and energy to develop a classifier for each variety of rice, deep transfer learning was firstly introduced to rice disease detection across different rice varieties. Three deep transfer learning methods were adapted for 12 transfer tasks, namely, fine-tuning, deep CORrelation ALignment (CORAL), and deep domain confusion (DDC). A self-designed convolutional neural network (CNN) was set as the basic network of the deep transfer learning methods. Fine-tuning achieved the best transferable performance with an accuracy of over 88% for the test set of the target domain in the majority of transfer tasks. Deep CORAL obtained an accuracy of over 80% in four of all the transfer tasks, which was superior to that of DDC. A multi-task transfer strategy has been explored with good results, indicating the potential of both pair-wise, and multi-task transfers. A saliency map was used for the visualization of the key wavelength range captured by CNN with and without transfer learning. The results indicated that the wavelength range with and without transfer learning was overlapped to some extent. Overall, the results suggested that deep transfer learning methods could perform rice disease detection across different rice varieties. Hyperspectral imaging, in combination with the deep transfer learning method, is a promising possibility for the efficient and cost-saving field detection of rice diseases among different rice varieties.

show abstract

Transfer‐learning‐based Raman spectra identification

Cited by 57 publications

References 31 publications

A novel polynomial reconstruction algorithm‐based 1D convolutional neural network used for transfer learning in Raman spectroscopy application

A novel polynomial reconstruction algorithm‐based 1D convolutional neural network used for transfer learning in Raman spectroscopy application

Pseudotargeted Metabolomic Fingerprinting and Deep Learning for Identification and Visualization of Common Pathogens

Hyperspectral Imaging Combined With Deep Transfer Learning for Rice Disease Detection

Contact Info

Product

Resources

About