We present Segment Anything Model 2 (SAM 2), a foundation model towards solving promptable visual segmentation in images and videos. We build a data engine, which improves model and data via user interaction, to collect the largest video segmentation dataset to date. Our model is a simple transformer architecture with streaming memory for real-time video processing. SAM 2 trained on our data provides strong performance across a wide range of tasks. In video segmentation, we observe better accuracy, using 3x fewer interactions than prior approaches. In image segmentation, our model is more accurate and 6x faster than the Segment Anything Model (SAM). We believe that our data, model, and insights will serve as a significant milestone for video segmentation and related perception tasks.

In the past couple of years, the landscape of image and video generation has transformed dramatically. Despite such phenomenal progress, rigorous and holistic evaluation of the generative models continues to suffer. This is primarily due to the multi-faceted and highly subjective nature of the task: the generated image / video should be evaluated not just on overall visual quality and aesthetics, but also on its alignment to the input prompt , originality, lack of propagating stereotypical biases, and several more factors.

In this talk, I’ll give an overview of current metrics, their shortcomings, and the rapid progress in the research community to improve the rigor in evaluation.

EfficientSAM: small but mighty.

Viewer experience is a complex outcome of many competing dimensions, from encoding quality to network performance of the video pipeline. Metrics, like Zero Buffer Rates (ZBR), illuminate the impact of various components in the end-to-end pipeline on the viewer experience. . This talk will explore the essential elements of operational excellence in video infrastructure, working backwards from viewer’s perspective and into the video pipeline.

Presentation information coming soon!

While video input pre-filtering is not a new concept, it has evolved significantly in recent years. Originally designed to reduce noise and artifacts in low-quality video, pre-filtering techniques have now been adapted for use even on pristine video content. This talk will explore the importance of video input pre-filtering and the specific advantages it can offer in modern video production and delivery workflows.

Key benefits of advanced input pre-filtering include reduced file sizes without sacrificing perceived video quality, as well as mitigation of common video quality issues like softness and deblocking artifacts. However, implementing a production-ready, broadcast-grade pre-filtering solution requires careful consideration of several critical factors.

This talk will dive into the core pillars of the newly released video input pre-filter in AWS Elemental's media processing solutions. It will explain how this advanced pre-filtering technology can help video providers deliver high-quality, efficiently encoded video for a wide range of applications, from over-the-top streaming to broadcast television. Attendees will come away with a deep understanding of the evolving role of pre-filtering in the video technology landscape and practical insights they can apply to their own workflows.