Nara · AI models
AI models bundled with Nara
Nara’s model manager ships more than 60 model families, downloaded and checksum-verified, then loaded on demand. Custom models from Reva deploy with no configuration, third-party ONNX models run where compatible nodes exist, and promptable models detect or segment objects described in words, with no training.
22 capabilities · model pack 261018 · example models for technical readers
- 01
General object detection
Detects common objects such as people, vehicles, animals, bags and phones.
Used in Item and pallet counting, Robot and AMR orchestration, Traffic monitoring, Traffic violation detection, Intrusion detection
General object detection pack
- 02
People detection
Detects people robustly in crowded or difficult scenes.
Used in PPE detection, Danger zone monitoring, People counting, Heatmaps and dwell time, Staff presence monitoring, Queue monitoring, Fall detection, Intrusion detection, Occupancy-based energy control
PeopleNet, RT-DETRv4
- 03
Object tracking and analytics
Tracks objects across frames for line-crossing counts, zone counts, dwell time, heatmaps and abandoned objects.
Used in Item and pallet counting, Danger zone monitoring, People counting, Heatmaps and dwell time, Queue monitoring, Traffic monitoring, Traffic violation detection, Intrusion detection
Built-in tracker
- 04
Person re-identification (ReID)
Recognises the same person across cameras and over time, without faces.
Used in People counting, Staff presence monitoring, Intrusion detection
FastReID, PersonViT, ReIdentificationNet
- 05
Face detection and recognition
Finds faces and matches them against a known-face gallery.
Used in Face recognition access control
RetinaFace, SCRFD, YuNet, ArcFace, EdgeFace, SFace
- 06
Face anti-spoofing and quality
Rejects photos and screens, and scores face image quality.
Used in Face recognition access control
MiniFASNetV2, eDifFIQA
- 07
Facial expression
Classifies expressions such as happy, sad, angry and surprised.
Used in Visitor demographics
AffectNet-7, OpenCV FER
- 08
Age and gender estimation
Estimates demographics for anonymous analytics.
Used in Visitor demographics
MiVOLO, UniFace
- 09
Pose estimation
Body keypoints for posture and gesture.
Used in Fall detection
RTMPose, RTMO, MoveNet
- 10
Head pose and gaze
Where a person is looking, for attention and engagement.
Used in Heatmaps and dwell time
6DRepNet360, UniFace HeadPose
- 11
Action and behaviour recognition
Falls, fights or violence, and other actions from pose or video.
Used in Fall detection
ST-GCN Fall, PoseC3D, SlowFast, VideoMAE
- 12
Crowd counting
Estimates people counts in dense crowds.
PET crowd point
- 13
Fire and smoke detection
Detects flames and smoke in camera views.
Used in Fire and smoke detection
Fire/smoke detection pack
- 14
PPE detection
Helmets, vests, masks, hairnets, gloves, coats and other PPE on each person.
Nabrio PPE models
- 15
Weapon detection
Firearms and knives, with tracking of the people involved.
Nabrio weapon models
- 16
Vehicle make and model recognition
Identifies a car’s brand and model.
Used in Licence plate recognition
Nabrio vehicle models
- 17
Licence plate detection
Finds number plates for recognition.
Used in Traffic violation detection, Licence plate recognition
Number plate detection pack
- 18
OCR (text reading)
Reads printed text, including Thai.
Used in Visual quality inspection, Sorting and grading, Gauge reading, Machine status monitoring, Barcode and label inspection, Traffic violation detection, Licence plate recognition
PP-OCRv5 / v6, CRNN, DB
- 19
Barcode and QR code reading
Reads 1D and 2D codes.
Used in Visual quality inspection, Sorting and grading, Barcode and label inspection
Built-in code reader
- 20
Promptable detection and segmentation
Finds or outlines anything described in words, with no training.
LocateAnything 3B, SAM 3.1
- 21
Image similarity and matching
Compares images or finds a reference object.
DINOv3, LightGlue
- 22
AI decisions (language and vision-language)
Answers custom questions about text, images and video, such as “Is the work area tidy?”
Decider, OmniJev
[ 02 ]Beyond deep learning
Classic vision tools
Not every task needs deep learning. Nara also includes colour detection and thresholds, circle and contour detection, feature matching, crop, rotate, perspective correction, brightness, contrast and white balance.
[ 03 ]Node catalogue
About 160 nodes in Nara 2.0
Grouped as in the editor. The full reference is on help.nabrio.com.
- TriggersOn deploy, time interval, recurring schedule, dashboard action, webhook, WebSocket message
- I/OIP camera (single and multi), Basler, webcam, media files; Modbus read/write; MQTT publish; HTTP request; TCP/IP client; email; Microsoft Teams; Telegram; Patlite and Primus tower lights; export image or text
- Vision / AIObject, people, face, pose, fire, plate, OCR and code nodes; trackers with line and zone counters, dwell time, heatmap and abandoned-object detection; ReID; promptable detection and segmentation; auto-labelling
- ProcessMaths, rounding, moving-window statistics, JSON and XML parsing, arrays and objects, text templates, time deltas, the AI Decision node
- OperationIf, switch/case, for and while loops, count-to, delay, stop, functions (series or parallel), procedure order, Python script
- DataVariables and arrays, database tables (create, read, filter, insert, modify, delete)
Advanced industrial nodes, such as industrial cameras, PLC and robot connectivity, are Professional Nodes, and their availability depends on your tier.
Full node referenceSee what Nara sees, on your own cameras.
Download Nara and build your first flow in minutes, or tell us what you need to see and do, and we’ll show you how it runs on your own hardware.