A Closer Look at Weakly-Supervised Audio-Visual Source Localization (Supplementary Material) Shentong Mo Carnegie Mellon University Pedro Morgado University of Wisconsin-Madison

Neural Information Processing Systems 

Code is available at: https://github.com/stoneMo/SLAVC. We conducted a comprehensive benchmarking study of existing approaches. Since we're interested in assessing the model's performance both when sounding objects are present We thus ignore the confidence threshold ( i.e ., We use this metric when comparing to results reported in the original papers. A verage precision (AP) is another metric often used in object detection. We compute the full curve (without interpolation).