Research Assistant

Texas A&M University, College Station

Unveiling the Future of Image Analysis: Introducing the Neighborhood Similarity Feature Selection Layer

Conducting research on 'Neighborhood Similarity Feature Space - NSFS (Computer Vision)' under Dr. Joshua Peeples' supervision, collaborating with the Los Alamos National Laboratory to develop a new feature extraction layer based on neighborhood similarity.

Skills: Computer Vision, Image Processing, PyTorch, Deep Learning, Machine Learning, Transformers, Image Retrieval

In January 2024, I embarked on an exciting journey with the Advanced Vision and Learning Lab (AVLL) at Texas A&M University, a hub dedicated to pioneering advancements in artificial intelligence (AI), machine learning (ML), and computer vision (CV). At AVLL, our mission transcends mere innovation; we strive to redefine the boundaries of AI/ML/CV research through a blend of ingenuity, diversity, and excellence.

Within this dynamic environment, I am currently immersed in the development of a groundbreaking feature extraction layer known as the Neighborhood Similarity Feature Selection (NSFS) layer. While conventional convolutional layers excel in capturing rich appearance information within images, they often overlook crucial structural details. This oversight stems from the inability of learned convolutional weights to explicitly represent neighborhood relationships.

To address this limitation, our NSFS layer draws inspiration from Local Binary Patterns (LBP) and is designed to encode intricate structural information by representing the relationships between central pixels and their neighbors. Each channel of the NSFS layer encapsulates the similarity or dissimilarity measures between the central point and one of its neighboring pixels, thereby enhancing the model's ability to discern nuanced spatial arrangements and orientations within objects.

My research objectives encompass a multifaceted exploration, including the identification of optimal similarity/dissimilarity measures, the rigorous evaluation of method robustness, and the delineation of precise task definitions for the NSFS layer. Additionally, I am actively engaged in the integration of learnable parameters to further enhance the adaptability and effectiveness of the NSFS layer across diverse applications.

Through this endeavor, I have honed a multitude of skills ranging from advanced image processing techniques to algorithm development and rigorous experimental evaluation. Moreover, my journey at AVLL has instilled in me a deep appreciation for collaborative research, intellectual curiosity, and the relentless pursuit of excellence.