Roboflow Deploy

Your custom model, everywhere

Once you've trained a model, you can get predictions wherever you need them without touching your model architecture. Your model runs everywhere you need it, automatically.

pip install Roboflow

Automate and integrate any part of the Roboflow workflow into your codebase

Programmatically export data in multiple formats from your dataset or upload images to grow your dataset

Perform inference with specific model versions and save predictions that you've made on the model

Collect sample inferences whether at specific time intervals, at random, and with different confidence thresholds to improve your model's performance

Deploy to the Cloud

Your computer vision pipeline is automatically set up to handle demanding workloads



Hosted API to receive predictions from your model as fast as your network connection can handle sending the frames



Autoscaling infrastructure includes load balancing, supports burst and no burst, and is always on without any custom engineering work



Gather multiple data points from each piece of data to be use in applications or create insights in real-time

Deploy to the Edge

Run your model on embedded devices for drones, robotics, IoT, offline scenarios, and more

Use Luxonis OAK devices via pip install roboflowoak or via Docker for a standardized device that combines a camera with a built in hardware accelerator

Optimized it to get maximum performance from the NVIDIA Jetson line of edge-AI devices by specifically tailoring the drivers, libraries, and binaries specifically to its CPU and GPU architectures

Mobile iOS SDK is available for deploying your model directly to an iOS application running on iPhone or iPad

Deploy to the Browser

Seamlessly add computer vision functionality into web applications

Roboflow.js is a custom layer on top of Tensorflow.js to enable real-time inference via Javascript at 30+ FPS

A browser-based way to quickly test model predictions and share your model without a new environment to host a model, save weights, or any GPU hassle

Customize to match your brand, add logos, remove buttons, and add text. The source code for the HTML, CSS, and JavaScript for these pages can be hosted anywhere.

Self-hosting and Enterprise Deployments

Deploy our Docker container in your private cloud (or on your own metal) using our Inference Server and our License Server Docker containers.



Run using CPU on a self-contained server in multiple locations to ensure your models are firewalled and able to run independently for maximum security



Ensure that sensitive images never leave your private network using Inference Server to receive requests from client applications inside the private network, with no Internet connection



Configure the Roboflow Inference Server to cache weights for up to 30 days to run completely air-gapped or in locations where an Internet connection is not readily available



Install NVIDIA drivers and nvidia-container-runtime, allowing Docker to pass through your GPU to the inference server

Your Models Improve Over Time

Use active learning to improve your models using real-world production data



Programmatically sample data once you identify problems or areas for improvement with your model



Cache data on your deployment device or system for a future upload to Roboflow for annotation



Your model will quickly learn what to detect and what not to detect with increasing confidence in the environments it is present in

Deploy custom object detection models in minutes