Back to Portfolio
Project

SeeSharp

Real-time video super resolution with PyTorch

Overview

PyTorch implementation of ERSVR (Enhanced Real-time Video Super Resolution) network using recurrent multi-branch dilated convolutions for 4x video upscaling.

Key Features

  • Multi-Branch Dilated Convolution module for efficient feature extraction
  • Feature Alignment Block enabling temporal consistency across video frames
  • Student-Teacher Knowledge Distillation achieving 101KB model size
  • Real-time performance with sharper edges and fine details compared to bicubic interpolation
  • Multiple testing interfaces: CLI, Web (Flask), and Programmatic API

Technologies Used

PythonPyTorchOpenCVFlaskComputer VisionKnowledge DistillationNeural Networks