Leveraging pre-trained models to accelerate learning on new datasets, we demonstrate significant improvements using transfer learning techniques.

We systematically evaluate various data augmentation techniques to improve generalization of convolutional neural networks on challenging image datasets.

Introducing semi-supervised learning methods for visual recognition tasks, combining limited labeled data with large unlabeled datasets to improve model performance.

This work proposes a set of convolutional models optimized for mobile devices without compromising accuracy in standard image classification tasks.

We explore several state-of-the-art neural network architectures for image recognition and present performance benchmarks across multiple datasets.