Deep learning Neural Networks have been a subject of interest for the wide range of problems they
can be effectively applied too. However, current architectures are unable to operate on very large
datasets such as high-resolution medical microscopy due to the computational and memory
requirements. A common approach to tackle this problem is to reduce the image resolution, which
leads to loss of massive information. The information that can be critical for the classification task.
Models which effectively use high resolution images are essential for applications in the fields of
histopathology, autonomous vehicles, and many others. In this work, we propose exploring various
novel methods and evaluate them for their performance on high-resolution data. We demonstrate
that by deploying patches-base attention deep neural network, we can leverage the importance of
image patches to improve the performance and decrease the overall computation.
This publication is available in alternative media on request. Penn State is an equal opportunity, affirmative action employer, and is committed to providing employment opportunities to all qualified applicants without regard to race, color, religion, age, sex, sexual orientation, gender identity, national origin, disability or protected veteran status.