Question about Roboflow inference credit

Hi, I want to ask about the 1000 free monthly inference credits. How credits deducted when I do the following:

  1. Iterate through a video and infer every frame of the video (I call the API for the amount of frames the video have)
  2. Infer the video directly
    Any help is appreciated!

Hi @chymchym1905 - when you infer frame-by-frame, each frame counts as one inference. Anything over 1,000 inference costs $0.003 per API call.

If you use our hosted video API, you are metered on the number of hours of footage you send. Free plans get 5 hours of video per month (see more on the pricing page).

1 Like

Thank you! I also want to ask if I can run video inference using a classification model at all? Im getting this error when im trying to use predict_video with my deployed model.

AttributeError: 'ClassificationModel' object has no attribute 'predict_video'

This is the code used for inference:

job_id, signed_url, expire_time = model.predict(
    r"path/to/video.mp4",
    fps=30,
    prediction_type="batch-video",
)
results = model.poll_until_video_results(job_id)

Hello there! Our Python package does not presently support running classification models. I am working on a fix for this and will send an update when the change is published.