List All Projects List Graduate Projects List Senior Design Projects List REU Projects Other Undergraduate Projects
Project # 4
Study of Deep Neural Networks Trained with Salient, Compressed Medical Video Data for Enhanced Predication
| Team Lead: | Aileen Sengupta |
| Other Team Members: | |
| Faculty Advisor(s): | Dr. Venkat Devarajan |
| Department: | EE |
| Project Type: | Graduate Student |
| Title: | Study of Deep Neural Networks Trained with Salient, Compressed Medical Video Data for Enhanced Predication |
| Abstract: | Deep learning is a specialized branch of machine learning characterized by the use of multi-layered artificial neural networks. Image and video compression, which reduces file size while maintaining acceptable quality, is critical for efficient data storage, transmission and rendering. This thesis explores an approach that employs differentially compressed video sequence focusing on specific regions of interest to train a deep neural network in a way that prioritizes these targeted areas. By doing so, the network achieves enhanced prediction capabilities (EPC), evidenced by reduced storage requirements, decreased training time and comparable or improved predictive accuracy. Preliminary experiments support the feasibility of this EPC framework. The proposed method is further validated on a transformer model trained jointly on motion vectors and static frames, thereby testing the hypothesis that targeted compression strategies can maintain, or even improve, performance in deep learning applications. |
| Poster: | View Poster #4 (PDF) |
| Judges Scoring: | Submit Evaluation (judges only) |