In recent years, due to the shortage of timber resources, it has become necessary to reduce the excessive consumption of forest resources. Non-destructive testing technology can quickly find wood defects and effectively improve wood utilization. Deep learning has achieved significant results as one of the most commonly used methods in the detection of wood knots. However, compared with convolutional neural networks in other fields, the depth of deep learning models for the detection of wood knots is still very shallow. This is because the number of samples marked in the wood detection is too small, which limits the accuracy of the final prediction of the results. In this paper, ResNet-34 is combined with transfer learning, and a new TL-ResNet34 deep learning model with 35 convolution depths is proposed to detect wood knot defects. Among them, ResNet-34 is used as a feature extractor for wood knot defects. At the same time, a new method TL-ResNet34 is proposed, which combines ResNet-34 with transfer learning. After that, the wood knot defect dataset was applied to TL-ResNet34 for testing. The results show that the detection accuracy of the dataset trained by TL-ResNet34 is significantly higher than that of other methods. This shows that the final prediction accuracy of the detection of wood knot defects can be improved by TL-ResNet34.

Protein fold classification is a critical step in protein structure prediction. There are two possible ways to classify protein folds. One is through template-based fold assignment and the other is ab-initio prediction using machine learning algorithms. Combination of both solutions to improve the prediction accuracy was never explored before. We developed two algorithms, HH-fold and SVM-fold for protein fold classification. HH-fold is a template-based fold assignment algorithm using the HHsearch program. SVM-fold is a support vector machine-based ab-initio classification algorithm, in which a comprehensive set of features are extracted from three complementary sequence profiles. These two algorithms are then combined, resulting to the ensemble approach TA-fold. We performed a comprehensive assessment for the proposed methods by comparing with ab-initio methods and template-based threading methods on six benchmark datasets. An accuracy of 0.799 was achieved by TA-fold on the DD dataset that consists of proteins from 27 folds. This represents improvement of 5.4–11.7% over ab-initio methods. After updating this dataset to include more proteins in the same folds, the accuracy increased to 0.971. In addition, TA-fold achieved >0.9 accuracy on a large dataset consisting of 6451 proteins from 184 folds. Experiments on the LE dataset show that TA-fold consistently outperforms other threading methods at the family, superfamily and fold levels. The success of TA-fold is attributed to the combination of template-based fold assignment and ab-initio classification using features from complementary sequence profiles that contain rich evolution information. http://yanglab.nankai.edu.cn/TA-fold/ Supplementary data are available at Bioinformatics online.

Taxonomic studies of leaves are one of the most effective means of correctly identifying plant species. In this paper, mixed activation function is used to improve the ResNet50 network in order to further improve the accuracy of leaf recognition. Firstly, leaf images of 15 common tree species in northern China were collected from the Urban Forestry Demonstration Base of Northeast Forestry University (45°43′–45°44′ N, 126°37′–126°38′ E, forest type was artificial forest), and a small leaf dataset was established. After that, seven commonly used activation functions were selected to improve the ResNet50 network structure, and the improved network was applied to the transfer learning research of the leaf small dataset. On this basis, five activation functions with better performance were selected for the study of mixed activation functions in deep learning. Two of these five activation functions are arbitrarily selected for combination, and a total of twenty combinations are obtained. Further, the first activation function was used in each combination to replace the first ReLU function after all addition operations in the ResNet50 network residual block structure, and another activation function was used to replace the other position ReLU functions. The experimental results show that in the transfer learning of the leaf small dataset using the ResNet50 deep residual network, the appropriate combination of mixed activation functions can increase the performance of the improved network to a certain extent. Among them, the ELU-Swish1 combination has the most significant improvement effect on the network performance, whose final effective validation accuracy reaches 98.17%. Furthermore, the comparison with GoogLeNet and VGG-16 also demonstrates the excellent performance of the improved ELU-Swish1 ResNet50 (ES-ResNet50) network architecture. Finally, tests on the other two small leaf datasets, Flavia and Swedish, also demonstrate the performance improvement of ES-ResNet50. The validation accuracy of the improved ES-Resnet 50 algorithm on these two datasets reaches 99.30% and 99.39%, respectively. All these experiments prove that the recognition performance of leaf transfer learning using the ES-ResNet50 network is indeed improved, which may be caused by the complementarity of the e-exponential gradient of ELU and Swish1 activation functions in the negative region.

The precise segmentation of forest areas is essential for monitoring tasks related to forest exploration, extraction, and statistics. However, the effective and accurate segmentation of forest images will be affected by factors such as blurring and discontinuity of forest boundaries. Therefore, a Pyramid Feature Extraction-UNet network (PFE-UNet) based on traditional UNet is proposed to be applied to end-to-end forest image segmentation. Among them, the Pyramid Feature Extraction module (PFE) is introduced in the network transition layer, which obtains multi-scale forest image information through different receptive fields. The spatial attention module (SA) and the channel-wise attention module (CA) are applied to low-level feature maps and PFE feature maps, respectively, to highlight specific segmentation task features while fusing context information and suppressing irrelevant regions. The standard convolution block is replaced by a novel depthwise separable convolutional unit (DSC Unit), which not only reduces the computational cost but also prevents overfitting. This paper presents an extensive evaluation with the DeepGlobe dataset and a comparative analysis with several state-of-the-art networks. The experimental results show that the PFE-UNet network obtains an accuracy of 94.23% in handling the real-time forest image segmentation, which is significantly higher than other advanced networks. This means that the proposed PFE-UNet also provides a valuable reference for the precise segmentation of forest images.