Precision dairy farming technology is widely used to improve the management efficiency and reduce cost in large-scale dairy farms. Machine vision systems are non-contact technologies to obtain individual and behavioral information from animals. However, the accuracy of image-based individual identification of dairy cows is still inadequate, which limits the application of machine vision technologies in large-scale dairy farms. There are three key problems in dairy cattle identification based on images and biometrics: (1) the biometrics of different dairy cattle may be similar; (2) the complex shooting environment leads to the instability of image quality; and (3) for the end-to-end identification method, the identity of each cow corresponds to a pattern, and the increase in the number of cows will lead to a rapid increase in the number of outputs and parameters of the identification model. To solve the above problems, this paper proposes a cascaded dairy individual cow identification method based on DeepOtsu and EfficientNet, which can realize a breakthrough in dairy cow group identification accuracy and speed by binarization and cascaded classification of dairy cow body pattern images. The specific implementation steps of the proposed method are as follows. First, the YOLOX model was used to locate the trunk of the cow in the side-looking walking image to obtain the body pattern image, and then, the DeepOtsu model was used to binarize the body pattern image. After that, primary classification was carried out according to the proportion of black pixels in the binary image; then, for each subcategory obtained by the primary classification, the EfficientNet-B1 model was used for secondary classification to achieve accurate and rapid identification of dairy cows. A total of 11,800 side-looking walking images of 118 cows were used to construct the dataset; and the training set, validation set, and test set were constructed at a ratio of 5:3:2. The test results showed that the binarization segmentation accuracy of the body pattern image is 0.932, and the overall identification accuracy of the individual cow identification method is 0.985. The total processing time of a single image is 0.433 s. The proposed method outperforms the end-to-end dairy individual cow identification method in terms of efficiency and training speed. This study provides a new method for the identification of individual dairy cattle in large-scale dairy farms.

Few-shot fine-grained image classification (FSFGIC) aims to classify subspecies with similar appearances under conditions of very limited data. In this paper, we observe an interesting phenomenon: different types of image data augmentation techniques have varying effects on the performance of FSFGIC methods. This indicates that there may be biases in the features extracted from the input images. The bias of the acquired feature may cause deviation in the calculation of similarity, which is particularly detrimental to FSFGIC tasks characterized by low inter-class variation and high intra-class variation, thus affecting the classification accuracy. To address the problems mentioned, we propose an unbiased feature estimation network. The designed network has the capability to significantly optimize the quality of the obtained feature representations and effectively reduce the feature bias from input images. Furthermore, our proposed architecture can be easily integrated into any contextual training mechanism. Extensive experiments on the FSFGIC tasks demonstrate the effectiveness of the proposed algorithm, showing a notable improvement in classification accuracy.

Feature map reconstruction networks (FRN) have demonstrated significant potential by leveraging feature reconstruction. However, the typical process of FRN gives rise to two notable issues. First, FRN exhibits high sensitivity to noise, particularly ambient noise, which can lead to substantial reconstruction errors and hinder the network’s ability to extract meaningful features. Second, FRN is particularly vulnerable to changes in data distribution. Owing to the fine-grained nature of the training data, the model is highly susceptible to overfitting, which may compromise its ability to extract effective feature representations when confronted with new classes. To address these challenges, this paper proposes a novel main feature selection module (MFSM), which suppresses feature noise interference and enhances the discriminative capacity of feature representations through principal component analysis (PCA). Extensive experiments validate the effectiveness of MFSM, revealing substantial improvements in classification accuracy for few-shot fine-grained image classification (FSFGIC) tasks.

To overcome the limitation of dynamic reciprocity, a new method for designing broadband on-chip optical isolators is proposed and demonstrated based on saturated gain, which is able to support simplex and duplex operation modes. By connecting a saturated gain waveguide to an appropriate linear loss waveguide, broadband isolation is predicted and proved theoretically through saturated gain-induced non-reciprocal transmission. The proposed isolator is numerically demonstrated with an operating band of 59 nm and an isolation ratio of −20 dB at the central wavelength of 1550 nm. It is noteworthy that when the current pump changes, the isolator still works well and keeps the high isolation ratio at a different input power. The footprint of the whole device is 465 μm × 0.35 μm which satisfies the requirement of photonic integrated circuits. The proposed isolator, with the combined advantages of compact footprint, broadband, duplex operation and high isolation, can enable on-chip unidirectional transmission and complex topological routing designation.

User interest mining on Sina Weibo is the basis of personalized recommendations, advertising, marketing promotions, and other tasks. Although great progress has been made in this area, previous studies have ignored the differences among users: the varied behaviors and habits that lead to unique user data characteristics. It is unreasonable to use a single strategy to mine interests from such varied user data. Therefore, this paper proposes an adaptive model for user interest mining based on a multi-agent system whose input includes self-descriptive user data, microblogs and correlations. This method has the ability to select the appropriate strategy based on each user's data characteristics. The experimental results show that the proposed method performs better than the baselines.

Few-shot fine-grained image classification aims to recognize subcategories with high visual similarity using only a limited number of annotated samples. Existing metric learning-based methods typically rely solely on spatial-domain features. Confined to this single perspective, models inevitably suffer from inherent texture biases, entangling essential structural details with high-frequency background noise. Furthermore, lacking cross-view geometric constraints, single-view metrics tend to overfit this noise, resulting in structural instability under few-shot conditions. To address these issues, this paper proposes the Dual-Subspace Network (DSNet). Specifically, DSNet utilizes the discrete cosine transform (DCT) and a low-pass filtering mechanism to explicitly isolate low-frequency global structural components from spatial features, thereby suppressing background interference. Truncated Singular Value Decomposition (SVD) is employed to construct independent, low-rank linear subspaces for both spatial texture and frequency structural features. An adaptive gating mechanism is designed to dynamically fuse the projection distances from these dual views. This strategy leverages the structural stability of the frequency subspace to prevent the spatial subspace from overfitting to background features. Extensive experiments on four benchmark datasets—CUB-200-2011, Stanford Cars, Stanford Dogs, and FGVC-Aircraft—demonstrate that DSNet exhibits excellent classification performance and robustness, achieving highly competitive results compared to existing metric learning algorithms. Complexity analysis further confirms that the proposed network achieves a favorable balance between high accuracy and computational efficiency, providing an effective new paradigm for few-shot fine-grained visual recognition.