Semi-supervised learning (SSL) has gained attention in medical image segmentation by leveraging abundant unlabeled data to reduce reliance on expert annotations. However, existing methods underutilize the complementary information from diverse augmented views. This information is crucial for handling the inherent variability and subtle pathological changes in medical imaging. To address this issue, we introduce Mixstill, a novel dual-component model comprising Mask-Distillation and Uncertainty Estimation. Mixstill enforces cross-view consistency on multiple data streams under diverse perturbations, promoting robust feature learning that enhances segmentation accuracy under anatomical variability and subtle pathological variations. Building upon this cross-view consistency, we further introduce an uncertainty-aware fusion mechanism that adaptively combines predictions from different views based on their reliability, producing high-quality supervision signals for unlabeled data. Extensive experiments conducted on two public datasets demonstrate the effectiveness of the proposed method. Under scarce labeled data conditions, our approach achieves superior performance compared to state-of-the-art SSL methods.

Knowledge Distillation (KD) transfers knowledge from a large pre-trained teacher network to a compact and efficient student network, making it suitable for deployment on resource-limited media terminals. However, traditional KD methods require balanced data to ensure robust training, which is often unavailable in practical applications. In such scenarios, a few head categories occupy a substantial proportion of examples. This imbalance biases the trained teacher network towards the head categories, resulting in severe performance degradation on the less represented tail categories for both the teacher and student networks. In this paper, we propose a novel framework called Knowledge Rectification Distillation (KRDistill) to address the imbalanced knowledge inherited in the teacher network through the incorporation of the balanced category priors. Furthermore, we rectify the biased predictions produced by the teacher network, particularly focusing on the tail categories. Consequently, the teacher network can provide balanced and accurate knowledge to train a reliable student network. Intensive experiments conducted on various long-tailed datasets demonstrate that our KRDistill can effectively train reliable student networks in realistic scenarios of data imbalance.

In recent years, computer-aided diagnosis (CAD) systems employing convolutional neural networks (CNNs) have achieved remarkable performance in medical image classification tasks. Despite this, deploying CNN-based CAD systems on medical equipment presents challenges due to their enormous computational and storage resource requirements. In this case, knowledge distillation reduces the cost of deploying CNNs by guiding a lightweight student network to learn from a robust teacher network. However, medical images have higher inter-class similarity than natural images, which makes it difficult for the teacher network to provide clear and accurate classification knowledge to the student network, resulting in the performance degradation of the student network. To address this problem, we divide the teacher predictions into clear predictions, ambiguous predictions, and misclassified predictions to analyze the interference caused by the similarity of medical images on knowledge distillation and propose a novel knowledge distillation frame-work, termed ClearKD. By enhancing ambiguous predictions and misclassified predictions with clear predictions as a reference, our ClearKD consistently provides high-quality teacher classification knowledge to the student network, increasing the ability of the student network to distinguish medical images. The experimental results on the skin lesions classification datasets (ISIC2019) and the brain tumor dataset demonstrate that our ClearKD outperforms existing state-of-the-art knowledge distillation methods in medical image classification tasks.

Multi-teacher knowledge distillation (MKD) aims to leverage the valuable and diverse knowledge presented by multiple teacher networks to improve the performance of the student network. Existing approaches typically rely on simple methods such as averaging the prediction logits or using sub-optimal weighting strategies to combine knowledge from multiple teachers. However, employing these techniques cannot fully reflect the importance of teachers and may even mislead student’s learning. To address these issues, we propose a novel Decoupled Multi-teacher Knowledge Distillation based on Entropy (DE-MKD). DE-MKD decomposes the vanilla KD loss and assigns weights to each teacher to reflect its importance based on the entropy of their predictions. Furthermore, we extend the proposed approach to distill the intermediate features from teachers to further improve the performance of the student network. Extensive experiments conducted on the publicly available CIFAR-100 image classification dataset demonstrate the effectiveness and flexibility of our proposed approach.

Knowledge distillation(KD) aims to obtain a lightweight student network with the target dataset's pre-trained network(s). In practical applications, the student network distilled on one dataset may fail to make fine-grained classifications of multiple categories(such as birds and dogs). To this end and to make better use of various datasets' pre-trained models, knowledge amalgamation (KA) strives to integrate the knowledge of multiple expert models trained on different datasets to attain a student network with multi-expert knowledge. Proposed KA methods for image classification ignore the problem that teacher networks may encounter with untrained class samples and provide misleading guidance to the student network. To address this problem, we propose a knowledge amalgamation framework based on uncertainty suppression. A series of experiments demonstrate the effectiveness of our framework; some of the experiments yield an accuracy improvement of 2% compared to the proposed methods.

Knowledge distillation aims to improve the generalization capacity of the student model by transferring knowledge from the teacher model. Existing feature-based methods explore knowledge transfer through hand-crafted feature mappings between teacher-student pairs. However, in different layers, the knowledge volume varies, and the knowledge exhibits semantic gaps. This leads to the possibility that hand-crafted layer associations may not enable the student model to effectively learn knowledge from the teacher model. We address this problem from two angles. On one hand, to ensure maximum knowledge transfer, we propose adaptive feature mapping based on the effective receptive field, which can quantify the knowledge volume of different layers and thus establish the optimal knowledge transfer paths between teacher-student pairs. On the other hand, to enhance the student model's ability to learn knowledge with semantic gaps from the teacher model, we propose adaptive feature fusion that fuses multiple intermediate layers of the teacher model as additional supervision. Experimental results demonstrate that the proposed method can significantly improve the performance of the student model.

Data-free knowledge distillation aims to learn a small student network from a large pre-trained teacher network without the aid of original training data. Recent works propose to gather alternative data from the Internet for training student network. In a more realistic scenario, the data on the Internet contains two types of label noise, namely: 1) closed-set label noise, where some examples belong to the known categories but are mislabeled; and 2) open-set label noise, where the true labels of some mislabeled examples are outside the known categories. However, the latter is largely ignored by existing works, leading to limited student network performance. Therefore, this paper proposes a novel data-free knowledge distillation paradigm by utilizing a webly-collected dataset under universal label noise, which means both closed-set and open-set label noise should be tackled. Specifically, we first split the collected noisy dataset into clean set, closed noisy set, and open noisy set based on the prediction uncertainty of various data types. For the closed-set noisy examples, their labels are refined by teacher network. Meanwhile, a noise-robust hybrid contrastive learning is performed on the clean set and refined closed noisy set to encourage student network to learn the categorical and instance knowledge inherited by teacher network. For the open-set noisy examples unexplored by previous work, we regard them as unlabeled and conduct self-supervised learning on them to enrich the supervision signal for student network. Intensive experimental results on image classification tasks demonstrate that our approach can achieve superior performance to state-of-the-art data-free knowledge distillation methods.

Text-guided image manipulation aims to use the given text description to modify the semantic content of the corresponding part in the input image. Although researchers have been obtained satisfactory performance in this field, they only 1) utilize the global sentence information at the initial modification stage and 2) exploit the fixed word information for regional adjustment in the subsequent modification process, hindering the improvement of image manipulation quality. Motivated by the mentioned issues, this paper proposes a novel approach to improve the performance of text-guided image manipulation by using sentence-aware and word-aware network. Concretely, we utilize global sentence information throughout the image manipulation process to improve the semantic consistency with the input text. On the other hand, we employ the dynamic selection method to dynamically adjust the word information corresponding to the regional image content to further improve the manipulation quality. As a result, our work surpasses the existing state-of-the-art methods on CUB and Oxford-102 flower datasets, demonstrating our effectiveness and superiority. In terms of Inception Score, our proposed method performs the most excellent performance. In terms of NIMA, the score of our method is closest to the score of the original dataset images, proving that our manipulated results are the most authentic.

Recent the-state-of-the-art image-based three-dimensional (3D) reconstruction methods that represent 3D shapes mainly using triangular mesh because of its memory efficiency and ability to present surface detail of objects compared to voxel and point cloud. Previous works usually follow an encoding and decoding pattern. A deep neural network to extract the features from the picture and reconstruct the 3D structure. It is a typical supervised learning process, requiring loss function to supervise the training. No existing works directly calculate the loss between the reconstruction mesh and ground truth mesh. Instead, they indirectly used the Chamfer Distance (CD) between point clouds as the loss. Most of the previous works focus on the encoding and decoding parts instead of the loss and CD is used for all works. However, when CD is applied to two point clouds with the same number of points, some points can match any number of points in another point cloud, so some points will be less involved in calculating the loss function, which will reduce the utilization of information. Therefore, We propose a new point matching strategy to calculate the loss. The point matching strategy we proposed limits the maximum number of matches for each point, allowing more points to be more involved in the loss calculation, thereby improving the information utilization rate. Experiments on single view reconstruction (SVR) and auto-encoding methods show that this new loss method can replace CD in this type of works and has better training results and 3D reconstruction quality.

Alzheimer’s disease (AD) is typical dementia, which is progressive and irreversible. Usually, the clinical diagnosis of patients is at a later stage, so early diagnosis can control the patient’s condition in time. The doctor is usually diagnosing the patient’s condition by 3D brain magnetic resonance imaging (MRI). However, because the 3D MRI structures of adjacent stages are almost similar, the multi-class diagnosis of AD becomes difficult. Therefore, there is a need to enhance the ability to extract more discriminative features from 3DMRI, promoting more accurate diagnosis. In addition, not only the entire MRI changes but also local areas in the MRI. Therefore, it is necessary to pay attention to changes in the entire image and local areas and to fuse features of different scales. In this paper, we propose an innovative convolutional network architecture for feature extraction and feature fusion. It consists of three modules: 1) a network based on ResNet-10, 2) 3D asymmetric convolution block (ACB), 3) multi-scale channel attentional feature fusion (MS-CAFF) module. The proposed model has been tested on the ADNI dataset and achieved an accuracy of 88.33%, which is nearly 2% higher than the latest research.

In this paper, we propose a novel drawing generative adversarial networks (DrawGAN) for text-to-image synthesis. The whole model divides the image synthesis into three stages by imitating the process of drawing. The first stage synthesizes the simple contour image based on the text description, the second stage generates the foreground image with detailed information, and the third stage synthesizes the final result. Through the step by step synthesis process from simple to complex and easy to difficult, the model can draw the corresponding results step by step and finally achieve the higher-quality image synthesis effect. Our method is validated on the Caltech-UCSD Birds 200 (CUB) dataset and the Microsoft Common Objects in Context (MS COCO) dataset. The experimental results demonstrate the effectiveness and superiority of our method. In terms of both subjective and objective evaluation, our method’s results surpass the existing state-of-the-art methods.

Convolutional neural networks(CNNs) have achieved a performance significantly superior to traditional machine learning methods. However, in the traditional machine learning methods, the feature extraction algorithms are compelling and beneficial for CNNs. This paper introduces the classic feature extraction algorithm gradient local binary pattern(GLBP) to the CNNs. More specially, the GLBP extractor weights will be fixed into the $3\times 3$ sized kernels to construct the GLBP layer to replace the first layer of CNNs. In the GLBP layer, the features extracted by the GLBP kernels will concate or add to the feature process by the convolutional kernels. Through extensive experiments, we demonstrated that the GLBP layer could efficiently improve CNNs performance. When training on the ImageNet dataset, the ResNet18 with GLBP layer obtained 1.19% Top-1 accuracy improvement and 0.87% Top-5 accuracy improvement, respectively.

In this paper, an Erudite Generative Adversarial Networks (EruditeGAN) is proposed for the text-to-image synthesis task. By introducing additional image distribution related to the original image into the network structure, the entire network can learn more about the image distribution and become more knowledgeable. In this case, it can be more clear about the distribution of the image that needs to be synthesized and finally synthesize high-quality results. Experiments well validate our method’s effectiveness and demonstrate the different effects of different distribution situations on the final results. According to the quantitative results of Fréchet Inception Distance (FID) and R-precision, our method’s comprehensive score is the best, which reflects our results are closer to the real image effect.

Knowledge distillation is a widely used neural network model compression technique. In general, the knowledge distillation transfer the knowledge from a large pre-trained teacher network with superior performance to a small student network enables the student network to achieve better performance. This paper proposes a triplet knowledge distillation framework (abbreviated as TKD), which introduces a smaller assistant network into the knowledge distillation structure. The performance of the assistant network is lower than that of the student network. During the training of the TKD, by minimizing the Mean Squared Error(MSE) loss function, the output of the student network will closer to the output of the teacher network and further from that of the assistant network. Therefore, the student network can learn more expressive knowledge from the teacher network while throwing away mistaken knowledge in the assistant network. Finally, the student network achieves a surprising performance even superior to the teacher network. We have demonstrated the effectiveness of TKD by extensive experiments on benchmark datasets(CIFAR-10, CIFAR-100, SVHN, STL-10). When using VGGNet as an experimental model, the student network VGGNet13 achieving 94.29%, 75.30%, 95.53%, and 87.61% accuracy on the CIFAR-10, CIFAR-100, SVHN, and STL-10 datasets, improved by 1.24%, 2.81%, 0.40%, and 2.32%, respectively.

In this paper, we propose a multi-perspective fusion method to improve the performance of text-to-image synthesis. From the perspective of the generator, we introduce a dynamic selection method to make the text feature match the corresponding image feature better, while the multi-class discriminant method with mask segmentation image as the extra type is introduced from the perspective of the discriminator to improve its discrimination ability. Through the effective integration of these two aspects of improvement, more excellent results by our method are obtained. Experiments on the Caltech-UCSD Birds 200 (CUB) and Microsoft Common Objects in Context (MS COCO) datasets demonstrate our method's effectiveness and superiority. The qualitative and quantitative experiments validate that our method is superior to the existing state-of-the-art methods.

Deep neural network technology is a milestone achievement in the field of computer vision. It obtained the performance that the shallow network cannot achieve through the multi-layer network structure and the learning method of reverse adjustment parameters. However, the feature extraction algorithm of the shallow network is very effective and also is more beneficial for deep neural networks. In this paper, we combine the shallow network algorithm to proposes the gradient local binary pattern layer(GLBP layer) to replace the first layer of Convolutional Neural Networks(CNNs). The GLBP layer plays a role in initializing the CNNs and can improve network performance without increasing the number and complexity of network layers. In the experiment, using the extracted layer modified by the GLBP feature algorithm to replace other classic deep neural networks, 2.65% and 2.9% performance improvements were obtained in the WideResNet16-2 and ResNet-101 respectively when training on CIFAR-100 dataset.

Convolutional neural networks(CNNs) are often over-parameterized and cannot apply to existing resource-limited artificial intelligence(AI) devices. Some methods are proposed to model compress the CNNs, but these methods are data-driven and often unable when lacking data. To solve this problem, in this paper, we propose a data-free model compression and acceleration method based on generative adversarial networks and network pruning(named DFNP), which can train a compact neural network only needs a pre-trained neural network. The DFNP consists of the source network, generator, and target network. First, the generator will generate the pseudo data under the supervise of the source network. Then the target network will get by pruning the source network and use these generated data for training. And the source network will transfer knowledge to the target network to promote the target network to achieve a similar performance of the source network. When the VGGNet- 19 is select as the source network, the target network trained by DFNP contains only 25% parameters and 65% calculations of the source network. Still, it retains 99.4% accuracy on the CIFAR-10 dataset without any real data.

Knowledge distillation is an extensively researched model compression technology, which uses a large teacher network to transmit information to a small student network. The critical point of the knowledge distillation method to improve the performance of the student network is to find an effective method to extract the information from the feature. The attention mechanism is a widely used feature processing method to process features effectively and obtain more expressive information. In this paper, we propose to use the dual attention mechanism in knowledge distillation to improve the performance of student networks, which extracts information from the spatial and channel dimensions of the feature. The channel dimension attention is search 'what' channel is more meaningful, and the spatial dimension attention is determine 'where' part of the feature is more expressive in a feature map. We have conducted extensive experiments on different datasets, shown that by implementing a dual attention mechanism to extract more expressive information for knowledge transfer, the student network can achieve performance beyond the teacher network.

With the rapid development of deep learning, convolutional neural networks(CNNs) have achieved great success. But these high-capability CNNs often with a huge burden of computation and memory, which hinders these CNNs from applying to practical application. To solve this problem, in this paper, we proposed a method to train a compact model with high-capacity. The student network with fewer parameters and calculations will learning from the knowledge of the teacher network with more parameters and calculations. To promote the ability of the student network, the more expressive knowledge is extracted from the middle-layer feature of neural networks by attention mechanism, and the knowledge transforms more effective from the teacher network to the student network by the positive- unlabeled(PU) classifier. We validate our method in extensive experiments, showing that it can train the student network to achieve significant performance superior to the teacher network.

A fundamental problem of text-to-image synthesis is the lack of quality assessment for a single generated image. Quantitative indicators of this work (such as Inception Score and Fréchet Inception Distance) only affect plenty of images' feature distribution. It causes monotonous evaluation and plenty of poor-quality image results. This paper proposes a new evaluation criterion for text-to-image synthesis by the blind image quality assessment(BIQA) method. To train the model, a Multi-Metrics Quality Assessment Dataset for generated birds' images(MMQA) is proposed. Besides, the Multi-hyper model is proposed to fit our dataset better. Experiments show that our method evaluates text- to-image tasks more comprehensively and optimize their results.

In recent years, local pattern based features have attracted increasing interest in object detection and recognition systems. Local Binary Pattern (LBP) feature is widely used in texture classification and face detection. But the original definition of LBP is not suitable for human detection. In this paper, we propose a novel feature set named gradient local binary patterns (GLBP), Original GLBP and Improved GLBP, for human detection. Experiments are performed on INRIA dataset, which shows the proposal GLBP feature is more discriminative than histogram of orientated gradient (HOG), histogram of template (HOT) and Semantic Local Binary Patterns (S-LBP), under the same training method. In our experiments, the window size is fixed. That means the performance can be improved by boosting and cascade methods. And the computation of GLBP feature is parallel, which make it easy for hardware acceleration. These factors make GLBP feature possible for real-time human detection.

In this paper, a feature named multi-scale bidirectional local template patterns (MBLTP) is proposed for human detection. As an extension of bidirectional local template patterns (BLTP), MBLTP not only integrates the textural and gradient information according to the four predefined templates but also calculates information for additional feature vectors by adjusting the scale of the training samples. These additional feature vectors contain multi-scale information on the samples, which can make the feature more discriminative than its original form. Experimental results for an INRIA dataset show that the detection rate of our proposed MBLTP feature outperforms those of other features such as the multi-level histogram of orientated gradient (multi-level HOG), multi scale block histogram of template (MB-HOT), and HOG-LBP. Moreover, in order to make our feature meet real-time requirements, an implementation based on a graphic process unit (GPU) is adopted to accelerate the calculation.