Objeck Documentation

All Bundles

Bundle ONNX Runtime inference library for running pre-trained models. Supports object detection (YOLO), image classification (ResNet), segmentation (DeepLab), pose estimation (OpenPose), face recognition, and Phi-3 text generation. Compile with -lib onnx.

FaceSession

Face recognition session wrapping an SCRFD detector and an ArcFace recognizer. Detect() returns bounding boxes and landmarks; Recognize() additionally extracts 512-dim embeddings for identity comparison via Compare(). Example session := FaceSession->New("det_10g.onnx", "w600k_r50.onnx"); r1 := session->Recognize(image1_bytes, 0.5); r2 := session->Recognize(image2_bytes, 0.5); if(r1->GetSize() > 0 & r2->GetSize() > 0) { faces1 := r1->GetResults(); faces2 := r2->GetResults(); sim := FaceSession->Compare(faces1[0]->GetEmbedding(), faces2[0]->GetEmbedding()); "Similarity: {$sim}"->PrintLine(); # >0.35 = same person }; session->Close();

Close #

Closes all loaded sessions and releases resources.

method : public : Close() ~ Nil

Compare # function

Compute cosine similarity between two L2-normalised ArcFace embeddings. Returns a value in [-1, 1]; typically >0.35 indicates the same person.

function : Compare(emb1:Float[], emb2:Float[]) ~ Float

Parameters

Name	Type	Description
emb1	Float	first embedding (512 floats)
emb2	Float	second embedding (512 floats)

Return

Type	Description
Float	cosine similarity score

Example

r1 := session->Recognize(image1, 0.5);
r2 := session->Recognize(image2, 0.5);
sim := FaceSession->Compare(r1->GetResults()[0]->GetEmbedding(), r2->GetResults()[0]->GetEmbedding());
"Same person? {$sim > 0.35}"->PrintLine();

Detect #

Detect all faces in an image.

method : public : Detect(image:Byte[], conf_threshold:Float) ~ FaceDetectionResult

Parameters

Name	Type	Description
image	Byte	raw image bytes (JPEG/PNG)
conf_threshold	Float	detection confidence threshold [0, 1]

Return

Type	Description
FaceDetectionResult	array of FaceDetection results

Example

session := FaceSession->New("det_10g.onnx");
image := FileReader->ReadBinaryFile("group.jpg");
result := session->Detect(image, 0.5);
"Faces found: {$result->GetSize()}"->PrintLine();
session->Close();

New # constructor

Constructor - detection only (no recognizer loaded).

New(det_model:String)

Parameters

Name	Type	Description
det_model	String	path to SCRFD detector ONNX model (e.g. det_10g.onnx)

New # constructor

Constructor - detection + recognition.

New(det_model:String, rec_model:String)

Parameters

Name	Type	Description
det_model	String	path to SCRFD detector ONNX model (e.g. det_10g.onnx)
rec_model	String	path to ArcFace recognizer ONNX model (e.g. w600k_r50.onnx)

New # constructor

Constructor with configuration options.

New(det_model:String, rec_model:String, config:Map<String,String>)

Parameters

Name	Type	Description
det_model	String	path to SCRFD detector ONNX model
rec_model	String	path to ArcFace recognizer ONNX model
config	Map<String,String>	key/value execution-provider options (e.g. "ep"->"cpu")

Recognize #

Detect faces and extract ArcFace embeddings for each. Requires the recognizer model to have been loaded.

method : public : Recognize(image:Byte[], conf_threshold:Float) ~ FaceRecognitionResult

Parameters

Name	Type	Description
image	Byte	raw image bytes (JPEG/PNG)
conf_threshold	Float	detection confidence threshold [0, 1]

Return

Type	Description
FaceRecognitionResult	array of FaceResult (detection + 512-dim embedding)

Example

session := FaceSession->New("det_10g.onnx", "w600k_r50.onnx");
image := FileReader->ReadBinaryFile("portrait.jpg");
result := session->Recognize(image, 0.5);
each(face in result->GetResults()) {
  "Face score: {$face->GetDetection()->GetScore()}"->PrintLine();
};
session->Close();