The complex backgrounds of crop disease images and the small contrast between the disease area and the background can easily cause confusion, which seriously affects the robustness and accuracy of apple disease- identification models. To solve the above problems, this paper proposes a Vision Transformer-based lightweight apple leaf disease- identification model, ConvViT, to extract effective features of crop disease spots to identify crop diseases. Our ConvViT includes convolutional structures and Transformer structures; the convolutional structure is used to extract the global features of the image, and the Transformer structure is used to obtain the local features of the disease region to help the CNN see better. The patch embedding method is improved to retain more edge information of the image and promote the information exchange between patches in the Transformer. The parameters and FLOPs (Floating Point Operations) of the model are significantly reduced by using depthwise separable convolution and linear-complexity multi-head attention operations. Experimental results on a complex background of a self-built apple leaf disease dataset show that ConvViT achieves comparable identification results (96.85%) with the current performance of the state-of-the-art Swin-Tiny. The parameters and FLOPs are only 32.7% and 21.7% of Swin-Tiny, and significantly ahead of MobilenetV3, Efficientnet-b0, and other models, which indicates that the proposed model is indeed an effective disease-identification model with practical application value.

Anti-cancer peptides (ACPs) are known as potential therapeutics for cancer. Due to their unique ability to target cancer cells without affecting healthy cells directly, they have been extensively studied. Many peptide-based drugs are currently evaluated in the preclinical and clinical trials. Accurate identification of ACPs has received considerable attention in recent years; as such, a number of machine learning-based methods for in silico identification of ACPs have been developed. These methods promote the research on the mechanism of ACPs therapeutics against cancer to some extent. There is a vast difference in these methods in terms of their training/testing datasets, machine learning algorithms, feature encoding schemes, feature selection methods and evaluation strategies used. Therefore, it is desirable to summarize the advantages and disadvantages of the existing methods, provide useful insights and suggestions for the development and improvement of novel computational tools to characterize and identify ACPs. With this in mind, we firstly comprehensively investigate 16 state-of-the-art predictors for ACPs in terms of their core algorithms, feature encoding schemes, performance evaluation metrics and webserver/software usability. Then, comprehensive performance assessment is conducted to evaluate the robustness and scalability of the existing predictors using a well-prepared benchmark dataset. We provide potential strategies for the model performance improvement. Moreover, we propose a novel ensemble learning framework, termed ACPredStackL, for the accurate identification of ACPs. ACPredStackL is developed based on the stacking ensemble strategy combined with SVM, Naïve Bayesian, lightGBM and KNN. Empirical benchmarking experiments against the state-of-the-art methods demonstrate that ACPredStackL achieves a comparative performance for predicting ACPs. The webserver and source code of ACPredStackL is freely available at http://bigdata.biocie.cn/ACPredStackL/ and https://github.com/liangxiaoq/ACPredStackL, respectively.

Natural breeding scenes have the characteristics of a large number of cows, complex lighting, and a complex background environment, which presents great difficulties for the detection of dairy cow estrus behavior. However, the existing research on cow estrus behavior detection works well in ideal environments with a small number of cows and has a low inference speed and accuracy in natural scenes. To improve the inference speed and accuracy of cow estrus behavior in natural scenes, this paper proposes a cow estrus behavior detection method based on the improved YOLOv5. By improving the YOLOv5 model, it has stronger detection ability for complex environments and multi-scale objects. First, the atrous spatial pyramid pooling (ASPP) module is employed to optimize the YOLOv5l network at multiple scales, which improves the model’s receptive field and ability to perceive global contextual multiscale information. Second, a cow estrus behavior detection model is constructed by combining the channel-attention mechanism and a deep-asymmetric-bottleneck module. Last, K-means clustering is performed to obtain new anchors and complete intersection over union (CIoU) is used to introduce the relative ratio between the predicted box of the cow mounting and the true box of the cow mounting to the regression box prediction function to improve the scale invariance of the model. Multiple cameras were installed in a natural breeding scene containing 200 cows to capture videos of cows mounting. A total of 2668 images were obtained from 115 videos of cow mounting events from the training set, and 675 images were obtained from 29 videos of cow mounting events from the test set. The training set is augmented by the mosaic method to increase the diversity of the dataset. The experimental results show that the average accuracy of the improved model was 94.3%, that the precision was 97.0%, and that the recall was 89.5%, which were higher than those of mainstream models such as YOLOv5, YOLOv3, and Faster R-CNN. The results of the ablation experiments show that ASPP, new anchors, C3SAB, and C3DAB designed in this study can improve the accuracy of the model by 5.9%. Furthermore, when the ASPP dilated convolution was set to (1,5,9,13) and the loss function was set to CIoU, the model had the highest accuracy. The class activation map function was utilized to visualize the model’s feature extraction results and to explain the model’s region of interest for cow images in natural scenes, which demonstrates the effectiveness of the model. Therefore, the model proposed in this study can improve the accuracy of the model for detecting cow estrus events. Additionally, the model’s inference speed was 71 frames per second (fps), which meets the requirements of fast and accurate detection of cow estrus events in natural scenes and all-weather conditions.

Sheep face recognition models deployed on edge devices require a good trade-off between model size and accuracy, but the existing recognition models cannot do so. To solve the above problems, this paper combines Mobilenetv2 with Vision Transformer to propose a balanced sheep face recognition model called MobileViTFace. MobileViTFace enhances the model’s ability to extract fine-grained features and suppress the interference of background information through Transformer to distinguish different sheep faces more effectively. Thus, it can distinguish different sheep faces more effectively. The recognition accuracy of 96.94% is obtained on a self-built dataset containing 5490 sheep face photos of 105 sheep, which is a 9.79% improvement compared with MobilenetV2, with only a small increase in Params (the number of parameters) and FLOPs (floating-point operations). Compared to models such as Swin-small, which currently performs SOTA, Params and FLOPs are reduced by nearly ten times, whereas recognition accuracy is only 0.64% lower. Deploying MobileViTFace on the Jetson Nano-based edge computing platform, real-time and accurate recognition results are obtained, which has implications for practical production.

It is an urgent task to improve the applicability of the cucumber disease classification model in greenhouse edge-intelligent devices. The energy consumption of disease diagnosis models designed based on deep learning methods is a key factor affecting its applicability. Based on this motivation, two methods of reducing the model’s calculation amount and changing the calculation method of feature extraction were used in this study to reduce the model’s calculation energy consumption, thereby prolonging the working time of greenhouse edge devices deployed with disease models. First, a cucumber disease dataset with complex backgrounds is constructed in this study. Second, the random data enhancement method is used to enhance data during model training. Third, the conventional feature extraction module, depthwise separable feature extraction module, and the squeeze-and-excitation module are the main modules for constructing the classification model. In addition, the strategies of channel expansion and = shortcut connection are used to further improve the model’s classification accuracy. Finally, the additive feature extraction method is used to reconstruct the proposed model. The experimental results show that the computational energy consumption of the adder cucumber disease classification model is reduced by 96.1% compared with the convolutional neural network of the same structure. In addition, the model size is only 0.479 MB, the calculation amount is 0.03 GFLOPs, and the classification accuracy of cucumber disease images with complex backgrounds is 89.1%. All results prove that our model has high applicability in cucumber greenhouse intelligent equipment.