Predictions

Use your deployed models to make predictions (inferences) on new data.

Make a prediction

Make a prediction using a model’s active version:

from seeme.types import ApplicationType

prediction = client.predict(
    model_id=my_model.id,
    item="path/to/image.jpg",
    application_type=ApplicationType.IMAGE_CLASSIFICATION
)

Parameter	Type	Description
model_id	str	The model id
item	str or dict	The input item (file path for images, text string for text, dict for structured data)
application_type	ApplicationType	The type of application. Default: `IMAGE_CLASSIFICATION`. Possible values: `IMAGE_CLASSIFICATION`, `OBJECT_DETECTION`, `TEXT_CLASSIFICATION`, `STRUCTURED`, `LANGUAGE_MODEL`, `NER`, `OCR`, `SPEECH_TO_TEXT`, `SEARCH`

The prediction result contains the following properties:

Property	Type	Description
id	str	Unique id for the inference
created_at	str	The creation date
updated_at	str	Last updated date
prediction	str	The predicted label (deprecated, use inference_items)
confidence	float	The prediction confidence (deprecated, use inference_items)
model_id	str	The model id used for prediction
model_version_id	str	The model version id used for prediction
extension	str	The file extension of the input
inference_items	List[InferenceItem]	List of prediction results
inference_time	str	Time taken for inference
result	str	Raw result as JSON string

Each InferenceItem has:

Property	Type	Description
prediction	str	The predicted label/class
confidence	float	The confidence score (0-1)
inference_id	str	The parent inference id
coordinates	str	Bounding box coordinates (for object detection)

Predict with a specific model version

Make a prediction using a specific model version instead of the active version:

prediction = client.version_predict(
    version=my_model_version,
    item="path/to/image.jpg",
    application_type=ApplicationType.IMAGE_CLASSIFICATION
)

Parameter	Type	Description
version	ModelVersion	The model version object
item	str or dict	The input item
application_type	ApplicationType	The type of application

Application-specific examples

Image Classification

prediction = client.predict(
    model_id=my_model.id,
    item="path/to/cat.jpg",
    application_type=ApplicationType.IMAGE_CLASSIFICATION
)

for item in prediction.inference_items:
    print(f"{item.prediction}: {item.confidence:.2%}")

Object Detection

prediction = client.predict(
    model_id=my_model.id,
    item="path/to/image.jpg",
    application_type=ApplicationType.OBJECT_DETECTION
)

for item in prediction.inference_items:
    print(f"{item.prediction} at {item.coordinates}")

Text Classification

prediction = client.predict(
    model_id=my_model.id,
    item="This movie was fantastic!",
    application_type=ApplicationType.TEXT_CLASSIFICATION
)

Structured/Tabular Data

input_data = {
    "feature1": 10,
    "feature2": "category_a",
    "feature3": 3.14
}

prediction = client.predict(
    model_id=my_model.id,
    item=input_data,
    application_type=ApplicationType.STRUCTURED
)

Get prediction history

Retrieve past predictions for a model:

inferences = client.get_inferences(
    model_id=my_model.id,
    model_version_ids=[],
    page_count=0,
    page_size=25,
    include_already_added=False,
    show_hidden=False
)

Parameter	Type	Description
model_id	str	The model id
model_version_ids	List[str]	Filter by specific model versions (empty for all)
page_count	int	Page number (0-indexed). Default: 0
page_size	int	Number of results per page. Default: 25
include_already_added	bool	Include inferences already added to datasets. Default: False
show_hidden	bool	Show hidden inferences. Default: False

Get prediction statistics

Get statistics about predictions made with a model:

stats = client.get_inference_stats(my_model.id)

Parameter	Type	Description
model_id	str	The model id

Jobs Share