This repository implements Robust Video Matting (RVM), a neural network specifically designed for robust human video segmentation. Unlike frame-by-frame approaches, RVM uses a recurrent architecture to leverage temporal information across video sequences, enabling high-quality matting without additional inputs. The model runs in real-time, achieving 4K at 76FPS and HD at 104FPS on an Nvidia GTX 1080 Ti GPU. Supports deployment across PyTorch, TensorFlow, TensorFlow.js, ONNX, and CoreML.