Fast Semantic Proposals for Image and Video Annotation using Modified Echo State Networks – Srikar Muppirisetty, Volvo Cars
Manually annotated data for semantic segmentation tasks is time consuming and tough to quality assure, accurate and automated region-based proposals can significantly aid high quality data annotation.
In this work, we propose a novel modified active-learning framework that iteratively learns from a small subset of data (20-30% images) and adapts to a variety of semantic segmentation goals without manual supervision on test images.
KEY TAKEAWAYS
We present a modified framework for Echo state networks to incorporate spatial and temporal information from neighbouring pixels and neighbouring images, respectively, to enable a variety of semantic segmentation goals.
Add comment