PredictionSession¶

class PredictionSession(url, input=None, image_id=None, video_id=None, dataset_id=None, project_id=None, api=None, tracking=None, tracking_config=None, **kwargs)[source]¶

Bases: object

Asynchronous inference session that yields Prediction.

The session starts inference immediately during construction and becomes an iterator. Use it directly when you need streaming results or progress control, or use higher-level helpers like predict_detached() method.

Exactly one of the following inputs must be provided: input, image_id, video_id, dataset_id, project_id (or their plural forms in kwargs: image_ids etc.).

Parameters:

url : str¶: Deployment base URL (e.g. https://.../net/<sessionToken>).
input : numpy.ndarray or str or PathLike or list, optional¶: Local input (NumPy image, image/video path, directory, or list of them).
image_id : int or List[int], optional¶: Image id (or list in image_ids kwarg).
video_id : int or List[int], optional¶: Video id (single video supported).
dataset_id : int or List[int], optional¶: Dataset id(s). Requires api to resolve datasets/projects.
project_id : int or List[int], optional¶: Project id (single project supported).
api=None¶: API client used for downloading by id / resolving datasets/projects.
tracking : bool, optional¶: Enable video tracking (if supported by deployment).
tracking_config : dict, optional¶: Tracking configuration dict (may include tracker key).
kwargs : dict¶: Additional inference settings (confidence, classes, batch size, etc.).

Raises:

AssertionError – If more than one input source is provided.
ValueError – For unsupported inputs or invalid combinations.

Usage Example:

import os
from dotenv import load_dotenv

import supervisely as sly
from supervisely.nn.model.prediction_session import PredictionSession

# Load secrets and create API object from .env file (recommended)
# Learn more here: https://developer.supervisely.com/getting-started/basics-of-authentication
if sly.is_development():
    load_dotenv(os.path.expanduser("~/supervisely.env"))

api = sly.Api.from_env()

session = PredictionSession(
    url='https://app.supervisely.com/net/<sessionToken>',
    image_id=123,
    api=api,
    conf=0.25,
)

for pred in session:
    _ = pred.boxes
    _ = pred.scores

Methods

`is_done`	Check whether server-side inference is finished.
`next`
`progress`	Return numeric progress of the current inference request.
`status`	Return raw status JSON for the current inference request.
`stop`	Stop the current inference request on the server (if running).