📄️ Classification
A classification pipeline returns a scalar result per class on each InferenceResult. See SimplifiedPipeline for the shared setup and Interpreting results for the result model.
📄️ Object detection
An object-detection pipeline returns a bounding_box result per detection on each InferenceResult. See SimplifiedPipeline for the shared setup and Drawing results to draw the detections.
📄️ Segmentation
A segmentation pipeline returns a segmentation result per detected region on each InferenceResult. Each carries class metadata, confidence, a bounding rectangle, and contour data. In C#, these are exposed as ClassLabel, Confidence, BoundingRect, and SubContours on SegmentationResult. See SimplifiedPipeline for the shared setup.
📄️ Instance segmentation
An instance-segmentation pipeline returns both a bounding_box and a segmentation on each InferenceResult. See SimplifiedPipeline for the shared setup.
📄️ Optical character recognition (OCR)
An OCR pipeline detects text regions and recognizes their content. Each detected region is a top-level InferenceResult with a boundingbox, and the recognized text is carried on its subresults. See SimplifiedPipeline for the shared setup.
📄️ Ready-to-use OCR package
The downloadable ready-to-use package is Python-specific. It includes the
📄️ Evaluating a sequence of images
A batched image tensor is evaluated in a single run, producing one ImageInferenceResult per image in input order. This extends the classification example to a sequence. See Running inference for the batching model.