Abstract
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 coarse-grain 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.
Highlights
An exponential growth of image classification has been witnessed over the few two decades and its semantic organization has become exceedingly expensive and difficult to manually categorize
In view of the issues above, we propose a learning method focusing on the problem of semi-supervised learning for fine-grained visual classification where we try to train a model that generalizes well on target samples, given the condition that there is a provision of both well-labeled source samples and labeled together with unlabeled target samples at training time
We explore each unit of Supervised Learning for Fine-Grained Classification (SSLGFC) on four publicly available fine-grained visual classification (FGVC) datasets
Summary
An exponential growth of image classification has been witnessed over the few two decades and its semantic organization has become exceedingly expensive and difficult to manually categorize. Thereby combining Convolutional Neural Network (CNN) with self-training becomes a powerful method that can learn a far better decision boundary and find a more advanced feature space that matches source and target samples distribution. This is somewhat similar to adversarial training based methods; it uses a simpler approach where feature learning is guided by a cross-entropy loss that enhances the closeness of the source and target features as well aligning the class-wise features. We obtained a great deal of performance on four of the most widely-used fine-grained recognition datasets, improving over some previous-best published methods
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
