Roboflow
ProductForumComputer Vision DatasetsComputer Vision Models
Search
⌃K
Links

Inference 2.0

Roboflow Inference and Deployment is getting an upgrade! Be the first to use it following these docs.
The Roboflow Inference API is an HTTP Roboflow API used for running inference with models trained on or uploaded to Roboflow.
Inference [Server] 2.0 is the preferred deployment method for supported Roboflow Deploy use cases. Model deployments not supported by Inference 2.0, are supported by Inference 1.0 (Legacy).

Model Deployment Options

  1. 1.
    Hosted Inference API
  2. 2.
    Hosted Inference Server
  3. 3.
    CPU (x86_64 CPU architectures)
  4. 4.
    NVIDIA Jetson (ARMv8 CPU architectures)
  5. 5.
    Raspberry Pi (ARMv8 CPU architectures)
  6. 6.
    Open AI CLIP
  7. 7.
    TensorRT (TRT) on NVIDIA GPU's (Roboflow Enterprise)
  8. 8.
    NVIDIA Jetson TensorRT (TRT) (Roboflow Enterprise)
  9. 9.
    Enterprise CPU (Roboflow Enterprise)
  10. 10.
    Enterprise NVIDIA GPU (Roboflow Enterprise)
  11. 11.
    Kubernetes

Model Deployment Options (Inference 1.0)

There are a number of deployment options available for Inference 1.0. The options for Object Detection are outlined below, with sections, by project or model type, linked to the appropriate documentation.
  1. 1.
    Hosted Inference API
  2. 2.
    Hosted Inference Server
  3. 3.
    CPU (x86_64 CPU architectures)
  4. 4.
    Web Browser (roboflow.js)
  5. 5.
    Raspberry Pi (ARMv8 CPU architectures)
  6. 6.
    Luxonis OAK
  7. 7.
    Mobile iOS SDK
  8. 8.
    TensorRT (TRT) on NVIDIA GPU's (Roboflow Enterprise)
  9. 9.
    NVIDIA Jetson TensorRT (TRT) (Roboflow Enterprise)
  10. 10.
    Enterprise CPU (Roboflow Enterprise)
  11. 11.
    Enterprise NVIDIA GPU (Roboflow Enterprise)
  12. 12.
    Kubernetes
Previous
Upload Weights
Next
Inference API Schema
Last modified 1mo ago