Run a Fine-Tuned Model

With Inference, you can run any of the 50,000+ models available on Roboflow Universe. You can also run private, fine-tuned models that you have trained or uploaded to Roboflow.

All models run on your own hardware.

Run a Model on Universe¶

In the first example, we showed how to run a rock paper scissors model. This model was hosted on Universe. Let's find another model to try.

Go to the Roboflow Universe homepage and use the search bar to find a model.

Browse the search page to find a model.

When you have found a model, click on the model card to learn more. Click the "Model" link in the sidebar to get the information you need to use the model.

Create a new Python file and add the following code:

import cv2
import inference
import supervision as sv

annotator = sv.BoxAnnotator()

def on_prediction(predictions, image):
    labels = [p["class"] for p in predictions["predictions"]]
    detections = sv.Detections.from_roboflow(predictions)
    cv2.imshow(
        "Prediction", 
        annotator.annotate(
            scene=image, 
            detections=detections,
            labels=labels
        )
    ),
    cv2.waitKey(1)

inference.Stream(
    source="webcam", # or rtsp stream or camera id
    model="coffee-cup-v2/3", # from Universe
    output_channel_order="BGR",
    use_main_thread=True, # for opencv display
    on_prediction=on_prediction, 
)

Replace coffee-cup-v2/3 with the model ID you found on Universe.

Then, run the Python script:

python app.py

Your webcam will open and you can see the model running:

Note: This model was tested on a Mac, but will achieve better performance on a GPU.

Run a Private, Fine-Tuned Model¶

You can run models you have trained privately on Roboflow with Inference. To do so, first go to your Roboflow dashboard. Then, choose the model you want to run.

Click the "Deploy" link in the sidebar to find the information you will need to use your model with Inference:

Copy the model ID on the page (in this case, taylor-swift-records/3).

Then, create a new Python file and add the following code:

import cv2
import inference
import supervision as sv

annotator = sv.BoxAnnotator()

def on_prediction(predictions, image):
    labels = [p["class"] for p in predictions["predictions"]]
    detections = sv.Detections.from_roboflow(predictions)
    detections = detections[detections.confidence > 0.9]
    print(detections)
    cv2.imshow(
        "Prediction", 
        annotator.annotate(
            scene=image, 
            detections=detections,
            labels=labels
        )
    ),
    cv2.waitKey(1)

inference.Stream(
    source="webcam", # or rtsp stream or camera id
    model="taylor-swift-records/3", # from Universe
    output_channel_order="BGR",
    use_main_thread=True, # for opencv display
    on_prediction=on_prediction, 
)

Replace taylor-swift-records/3 with the model ID from your private model.

Then, run the Python script:

python app.py

Your webcam will open and you can see the model running.