videoflow/videoflow

Python framework that facilitates the quick development of complex video analysis applications and other series-processing based applications in a multiprocessing environment.

What it solves

Videoflow is a framework for building distributed video and stream processing pipelines. It allows developers to define a processing graph (a directed acyclic graph or DAG) once and run it either locally for development or on Kubernetes for production, without changing the code.

How it works

Users describe their pipeline as a series of producers, processors, and consumers. Each node in the graph runs as an independent worker (a subprocess locally or a pod on Kubernetes) and communicates via a NATS JetStream message broker. Each node publishes its output to a specific broker subject and subscribes to its parents' subjects to reassemble inputs. To handle expensive setup like loading ML models, nodes use an open() method rather than the constructor.

Who it’s for

It is designed for developers building high-throughput video analytics pipelines that require scaling, GPU scheduling, and the ability to run complex processing chains across distributed environments.

Highlights

  • Dual Execution Modes: Run the same pipeline locally as subprocesses or on Kubernetes as containers with no code changes.
  • Distributed Architecture: Uses NATS JetStream for communication and Redis for payload storage.
  • C-level Scaling: Supports per-node scaling, GPU scheduling, and autoscaling on Kubernetes.
  • Language Agnostic: Can run components written in any language if they are shipped as container images.
  • Built-in Vision Components: Includes off-the-shelf components for object detection, tracking, pose estimation, and segmentation.
  • Robust Error Handling: Features a dead-letter queue (DLQ) for poison messages and automatic restarts for crashed nodes.

Related

  • Project
    nndeploy/nndeploy

    A high-performance AI deployment framework that uses visual workflows to deploy AI models across desktop, mobile, and edge devices.

  • Project
    OpenDCAI/DataFlow

    DataFlow is an open‑source Python framework that lets you build, run, and share low‑code pipelines for generating, cleaning, and evaluating LLM training data. It provides a library of reusable operators, a visual WebUI, an AI‑assistant for auto‑creating pipelines, and a Ray‑based distributed execution layer.

  • Project
    insight-platform/Savant

    A high-level framework for building real-time, high-performance computer vision and video analytics pipelines on Nvidia hardware, abstracting the complexity of DeepStream.

  • Project
    roboflow/roboflow-python

    Roboflow‑python is the official Python client for the Roboflow computer‑vision platform. It lets you programmatically create workspaces/projects, upload and version datasets, trigger model training, and run inference on hosted or self‑hosted models. Install via `pip install roboflow` (or the lightweight `roboflow-slim`), authenticate with an API key or login, and use a simple object hierarchy (workspace → project → version) to manage the full CV lifecycle.

  • Project
    daydreamlive/scope

    A tool for running and customizing real-time, interactive generative AI video pipelines using autoregressive diffusion models and a composable architecture.