MDS-A: New Dataset for Test-Time Adaptation in Object Detection

Discover the "Multiple Distribution Shift - Aerial" (MDS-A) dataset in this 17-minute presentation by ASU graduate student Noel Ngu at AAAI-MAKE Spring Symposium 2025. Learn about this innovative dataset specifically designed to better evaluate models for test-time adaptation in object detection. With its unique structure featuring multiple training and test sets across various distributions, MDS-A enables researchers to thoroughly investigate how object detection models perform when trained and tested across different distribution scenarios. Access the dataset through the provided website and explore the accompanying research paper published in the AAAI-MAKE/Spring Symposium 2025 proceedings. This presentation is part of the Neuro Symbolic Channel, which offers tutorials, courses, and research findings at the intersection of symbolic methods and deep learning, originally derived from Arizona State University's AI curriculum.