fastvideo/gpu-camera-sample

Image processing software on GPU (Windows, Linux, ARM) for real time machine vision camera applications. Performance benchmarks and Glass-to-Glass time measurements. MIPI CSI cameras support. RAW2RGB processing on CUDA with 16-bit ISP. Software for Jetson.

What it solves

This project provides a high-performance sample application for real-time GPU-accelerated image processing of raw camera data. It addresses the need for low-latency, high-throughput image signal processing (ISP) pipelines that can handle massive pixel rates (up to 4 GPix/s) without relying on fixed-function hardware ISPs, allowing for fully tunable processing stages.

How it works

The application implements a multithreaded pipeline on NVIDIA GPUs. It captures raw image data (monochrome or Bayer) from various industrial cameras via SDKs or GenICam, imports the data to the GPU, and executes a comprehensive ISP sequence including debayering, noise reduction, color correction, and gamma adjustment. The processed frames are then rendered via OpenGL or exported as compressed video streams (JPEG, H.264, HEVC) to SSD or network targets.

Who it’s for

It is designed for developers working with machine vision, industrial imaging, and robotics who use NVIDIA GPUs (including Jetson platforms) and need to implement custom, high-precision (16/32-bit) image processing pipelines for professional-grade cameras.

Highlights

  • Extreme Throughput: Capable of processing up to 4 GPix/s on high-end NVIDIA GPUs like the RTX 4090.
  • Comprehensive ISP Pipeline: Includes bad pixel correction, dark frame subtraction, flat-field correction, white balance, and multiple debayering algorithms (HQLI, L7, DFPD, MG).
  • Broad Camera Support: Compatible with XIMEA, FLIR, Basler, Lucid Vision Labs, and other GenICam-compliant cameras.
  • Low Latency: Includes built-in tools for measuring "glass-to-glass" latency from capture to display.
  • Cross-Platform: Supports Windows, Linux, and NVIDIA Jetson (ARM64) platforms.

Related

  • Project
    NVIDIA-NeMo/Curator

    A GPU-accelerated data curation framework for building repeatable pipelines to load, filter, deduplicate, and transform massive text, image, video, and audio datasets for AI training.

  • Project
    CVCUDA/CV-CUDA

    A GPU-accelerated library of computer vision algorithms designed to provide high-throughput, low-latency image and video processing for AI pipelines on NVIDIA platforms.

  • Project
    ros-perception/image_pipeline

    A ROS 2 package that processes raw camera images into a usable format for higher-level vision processing in robotics.

  • Project
    rapidsai/cucim

    cuCIM is a GPU-accelerated computer vision and image processing library designed for large, multidimensional images used in biomedical, geospatial, and remote sensing applications.

  • Project
    AaronFeng753/Waifu2x-Extension-GUI

    A graphical user interface for AI-powered image and video upscaling and frame interpolation, supporting multiple ML engines and hardware accelerators.