Optimization of DNN based computer vision algorithms for resource constrained tactical edge

The Army Research Laboratory Research Associateship Program (ARL-RAP) is offering a research opportunity focused on optimizing deep neural network-based computer vision algorithms for deployment on resource-constrained tactical edge computing devices. Object detection and semantic segmentation-based computer vision algorithms have become essential for visual scene understanding in tactical edge computing. However, these algorithms deploy complex deep neural network architectures that are computationally intensive and not optimized for edge devices with resource constraints. This research project involves profiling several computer vision algorithms and optimizing them for deployment on edge computing platforms. The selected candidate will develop a generalized model switching framework for dynamic loading of models to meet both resource constraints and mission requirements. The research will be conducted through the Computational and Information Sciences Directorate (CISD), which focuses on communications, atmospheric modeling, battlefield visualization, and computing. The participant will collaborate with both internal and external collaborators to advance project goals.