JIT-Masker: Efficient Online Distillation for Background Matting
This addresses the need for efficient virtual background tools for everyday consumers, though it is incremental as it builds on existing distillation and matting techniques.
The paper tackles real-time portrait matting for virtual backgrounds in video conferences by developing a pipeline that trades some accuracy for better throughput using online distillation, achieving a 5x speedup over a saliency-based method while maintaining higher quality.
We design a real-time portrait matting pipeline for everyday use, particularly for "virtual backgrounds" in video conferences. Existing segmentation and matting methods prioritize accuracy and quality over throughput and efficiency, and our pipeline enables trading off a controllable amount of accuracy for better throughput by leveraging online distillation on the input video stream. We construct our own dataset of simulated video calls in various scenarios, and show that our approach delivers a 5x speedup over a saliency detection based pipeline in a non-GPU accelerated setting while delivering higher quality results. We demonstrate that an online distillation approach can feasibly work as part of a general, consumer level product as a "virtual background" tool. Our public implementation is at https://github.com/josephch405/jit-masker.