Semi-supervised learning is a machine learning approach that tackles the challenge of having a large set of unlabeled data and few labeled ones. In this paper we adopt a semi-supervised self-training method to increase the amount of training data, prevent overfitting and improve the performance of deep models by proposing a novel selection algorithm that prevents mistake reinforcement which is a common thing in conventional self-training models. The model leverages, unlabeled data and specifically, after each training, we first generate pseudo-labels on the unlabeled set to be added to the labeled training samples. Next, we select the top-k most-confident pseudo-labeled images from each unlabeled class with their pseudo-labels and update the training data, and retrain the network on the updated training data. The method improves the accuracy in two-fold; bridging the gap in the appearance of visual objects, and enlarging the training set to meet the demands of deep models. We demonstrated the effectiveness of the model by conducting experiments on four state-of-the-art fine-grained datasets, which include Stanford Dogs, Stanford Cars, 102-Oxford flowers, and CUB-200-2011. We further evaluated the model on some coarsegrain data. Experimental results clearly show that our proposed framework has better performance than some previous works on the same data; the model obtained higher classification accuracy than most of the supervised learning models.

Traffic sign recognition is a classification problem that poses challenges for computer vision and machine learning algorithms. Although both computer vision and machine learning techniques have constantly been improved to solve this problem, the sudden rise in the number of unlabeled traffic signs has become even more challenging. Large data collation and labeling are tedious and expensive tasks that demand much time, expert knowledge, and fiscal resources to satisfy the hunger of deep neural networks. Aside from that, the problem of having unbalanced data also poses a greater challenge to computer vision and machine learning algorithms to achieve better performance. These problems raise the need to develop algorithms that can fully exploit a large amount of unlabeled data, use a small amount of labeled samples, and be robust to data imbalance to build an efficient and high-quality classifier. In this work, we propose a novel semi-supervised classification technique that is robust to small and unbalanced data. The framework integrates weakly-supervised learning and self-training with self-paced learning to generate attention maps to augment the training set and utilizes a novel pseudo-label generation and selection algorithm to generate and select pseudo-labeled samples. The method improves the performance by: (1) normalizing the class-wise confidence levels to prevent the model from ignoring hard-to-learn samples, thereby solving the imbalanced data problem; (2) jointly learning a model and optimizing pseudo-labels generated on unlabeled data; and (3) enlarging the training set to satisfy the hunger of deep learning models. Extensive evaluations on two public traffic sign recognition datasets demonstrate the effectiveness of the proposed technique and provide a potential solution for practical applications.

In hyperspectral image (HSI) classification, Convolutional Neural Networks (CNNs) have exhibited exceptional performance, owing to their hierarchical nonlinear modeling. However, their fixed square receptive field constrains their ability to effectively handle irregular image regions. Graph Convolution Networks (GCNs) have been introduced to learn irregular regions through correlations between adjacent pixels modeled as superpixel-based nodes, yet they lack pixel-level information. We propose a novel approach "Pixel-level with Covariance Pooling feature and Superpixel-level feature Fusion for Hyperspectral Image Classification" (PiCovS). Our method harnesses complementary spectral-spatial features at both pixel and superpixel levels to capture characteristics of both small-scale regular and large-scale irregular regions. We introduce a hybrid network that integrates and propagates features between image-level pixels and graph-level nodes using a graph encoder-decoder, effectively reconciling the differences between regular CNN and irregular GCN data representations. To enhance superpixel boundary learning, we modify the Manifold Simple Linear Iterative Clustering (M-SLIC) algorithm by incorporating texture feature information, resulting in refined superpixel representations. Additionally, we propose a novel covariance pooling mechanism with an attention mechanism within the CNN branch, enabling the capturing and utilization of holistic HSI information along spectral and spatial dimensions by exploiting second-order statistics throughout the network. Our comprehensive experiments showcase the efficiency and robustness of the proposed framework, achieving an impressive overall accuracy of 99.84%, 99.97%, 99.98%, and 81.96% on the Indian Pines, University of Pavia, Salinas, and the Houston University datasets, respectively. Remarkably, PiCovS excels even with limited training samples, outperforming other state-of-the-art methods in accuracy.

The unavailability of large amounts of well-labeled data poses a significant challenge in many medical imaging tasks. Even in the likelihood of having access to sufficient data, the process of accurately labeling the data is an arduous and time-consuming one, requiring expertise skills. Again, the issue of unbalanced data further compounds the abovementioned problems and presents a considerable challenge for many machine learning algorithms. In lieu of this, the ability to develop algorithms that can exploit large amounts of unlabeled data together with a small amount of labeled data, while demonstrating robustness to data imbalance, can offer promising prospects in building highly efficient classifiers. This work proposes a semisupervised learning method that integrates self-training and self-paced learning to generate and select pseudolabeled samples for classifying breast cancer histopathological images. A novel pseudolabel generation and selection algorithm is introduced in the learning scheme to generate and select highly confident pseudolabeled samples from both well-represented classes to less-represented classes. Such a learning approach improves the performance by jointly learning a model and optimizing the generation of pseudolabels on unlabeled-target data to augment the training data and retraining the model with the generated labels. A class balancing framework that normalizes the class-wise confidence scores is also proposed to prevent the model from ignoring samples from less represented classes (hard-to-learn samples), hence effectively handling the issue of data imbalance. Extensive experimental evaluation of the proposed method on the BreakHis dataset demonstrates the effectiveness of the proposed method.