studentTeacherAnomalyDetector

R2026b

Detect anomalies using Student-Teacher network

Since R2026b

Description

Add-On Required: This feature requires the Visual Inspection Toolbox Model for Student-Teacher Anomaly Detection add-on.

The studentTeacherAnomalyDetector object detects images of anomalies using a Student-Teacher anomaly detector network. Train the detector using the trainStudentTeacherAnomalyDetector function. To detect anomalous images, pass the trained detector to the classify function.

You can perform tile based anomaly detection using the Student-Teacher anomaly detector. You can train the detector on smaller tiles extracted from a larger image and then perform inference on the full image. For best results, the tiles and the full image must be of the same resolution.

Creation

Syntax

detector = studentTeacherAnomalyDetector

detector = studentTeacherAnomalyDetector(Name=Value)

Description

detector = studentTeacherAnomalyDetector creates a Student-Teacher anomaly detector.

detector = studentTeacherAnomalyDetector(Name=Value) sets writable properties using one or more name-value arguments. For example, Network="small" specifies use a small base network for the Student-Teacher anomaly detector network.

Name-Value Arguments

Threshold — Anomaly threshold

[] (default) | numeric scalar

Anomaly threshold, stored as a numeric scalar. You must set the value of the threshold before calling the classify function. You can calculate a threshold using the anomalyThreshold function.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

Network — Size of base network

"small" (default) | "medium"

Size of the base network, specified as "small" or "medium". This argument sets the Network property of the object.

Data Types: char | string

Output Arguments

detector — Student-Teacher anomaly detector

studentTeacherAnomalyDetector object

Student-Teacher anomaly detector, returned as a studentTeacherAnomalyDetector object.

Properties

Threshold — Anomaly threshold

[] (default) | numeric scalar

Anomaly threshold, specified as a numeric scalar. You must set the value of the threshold before calling the classify function or before exporting the network to ONNX with classification as one of the export tasks. You can calculate a threshold by using the anomalyThreshold function.

InputSize — Input image size

Read-only: [] (default) | three-element numeric row vector

This property is read-only.

Input image size, represented as a three-element numeric row vector of the form [rows columns channels]. The trainStudentTeacherAnomalyDetector function sets the InputSize property during training. The size of the images used to train the network determines the value of this property.

When performing tile-based anomaly detection, which consists of training the network on smaller tiles extracted from a larger image and then performing inference on the full image, ensure that the tiles and full image are of the same resolution.

Network — Size of base network

"small" (default) | "medium"

This property is read-only after object creation.

Size of the base network, represented as "small" or "medium".

Object Functions

predict	Predict unnormalized anomaly scores
classify	Classify image as normal or anomalous
anomalyMap	Predict per-pixel anomaly score map
exportONNXNetwork	Export Student-Teacher anomaly detector to ONNX model format

Examples

Create Student-Teacher Anomaly Detector

Create a studentTeacherAnomalyDetector object from a medium-sized base network.

detector = studentTeacherAnomalyDetector(Network="medium")

detector = 
  studentTeacherAnomalyDetector with properties:

    Threshold: []
    ImageSize: []
      Network: "medium"

More About

Student-Teacher Network Architecture

The Student-Teacher anomaly detector operates on the principle of a student-teacher model, where the teacher is a pretrained deep neural network that understands normal data behavior, and the student is a smaller network that learns to mimic the responses of the teacher on normal data. During training, you must use only normal data, without anomalies, enabling the student to learn a compact representation of what the teacher considers normal. At inference, both the student and teacher evaluate incoming data, detecting anomalies based on the discrepancy between the output of the student and teacher, as the student is likely to fail to accurately replicate the behavior of the teacher on anomalous data, highlighting instances that deviate from the learned normal behavior.

Extended Capabilities

C/C++ Code Generation

Generate C and C++ code using MATLAB® Coder™.

Usage notes and limitations:

• To prepare a studentTeacherAnomalyDetector object for code generation, use the coder.loadDeepLearningNetwork function.

GPU Code Generation

Generate CUDA® code for NVIDIA® GPUs using GPU Coder™.

Refer to the usage notes and limitations in the C/C++ Code Generation section. The same usage notes and limitations apply to GPU code generation.

Version History

Introduced in R2026b