Object detection
Put a box and a class on every instance in frame.
Applications
Computer vision that ships to real hardware
Train image classification and object detection models on your own data, then build a package for the device that has to run them.
- Classification and detection
- NanoVision · Vision Pro · MobileNetV2
- Ships to Raspberry Pi, UNO Q, TensorFlow Lite
What you get
What you get in the platform
Every stage of the workflow lives in one place — collecting and labelling images, designing the impulse, training, testing what came out, and shipping it to the device that has to run it.
- Data
- Train
- Evaluate
- Deploy
- Operate
Dataset managementData
Image labelingData
AI-assisted labelingData
Bounding boxesData
Dataset versionsData
Version controlData
ProjectsData
Users and accessData
Supported formatsData
Industrial camerasData
Impulse designTrain
Feature generationTrain
Vision Pro trainingTrain
NanoVisionTrain
Transfer learningTrain
GPU trainingTrain
Cloud trainingTrain
EvaluationEvaluate
Confusion matrixEvaluate
Model testingEvaluate
Model optimizationDeploy
Edge deploymentDeploy
OTA updatesDeploy
Fleet managementDeploy
Raspberry PiDeploy
Arduino UNO QDeploy
PXE runtimeDeploy
Device monitoringOperate
Inference logsOperate
AnalyticsOperate
REST APIOperate
Python SDKOperate
Labeling
AI-assisted Labeling
Accelerate dataset creation with AI-assisted image labeling. Automatically detect and suggest bounding boxes for objects, significantly reducing manual annotation time while improving consistency across large datasets. Review, adjust, and approve predictions in seconds to build high-quality training data faster.
- Automatic object suggestions
- Faster annotation workflow
- Human review & correction
- High-quality training datasets
How it works
The whole pipeline, one stage at a time
Every stage below is a screen in the platform, not a diagram of something you still have to build.
0110
01Stage 1 of 10
Collect data
Upload samples or import an existing folder tree.
Data
Key features
Built for image data, specifically
Bring your existing annotations
Import COCO, YOLO, Pascal VOC, Open Images and Edge Impulse JSON without reformatting them first.
Bounding-box annotation editor
Draw, adjust and relabel boxes directly on the dataset grid.
AI-assisted labeling
Describe what you are looking for in plain text and review the proposed boxes before they land.
NanoVision and Vision Pro detection
Centroid detection with NanoVision v1 or v2, or box detection with Vision Pro.
Image classification
Train MobileNetV2 from scratch or fine-tune with ImageNet transfer learning.
Deployment builds
Turn any trained model into a ready-to-run package for the target you pick.
Use cases
What teams build with it
All of these start the same way: your images, your classes, your device. Nothing here is a stock model — it is what the same workflow produces when you point it at a different problem.
People counting
Count how many people pass a fixed camera.
PPE detection
Check the required gear is worn before work starts.
Defect detection
Flag the parts that fail, and show where they fail.
Surface inspection
Grade finish and texture straight off the line.
Barcode reading
Locate codes on moving stock so a reader can lock on.
Text region OCR
Find the printed regions a reader needs to parse.
Quality inspection
Sort output into the pass and fail classes you track.
Retail analytics
See how shoppers move through a floor plan.
Vehicle detection
Pick out vehicles arriving at a yard or gate.
Parking monitoring
Track which bays are free, bay by bay.
Waste sorting
Separate recyclable material on a sorting belt.
Agriculture
Spot crop stress from a camera out in the field.
Wildlife monitoring
Identify species on trail cameras with no uplink.
Smart factory
Watch a cell and react when its state changes.
Production counting
Count finished units without a mechanical trigger.
Package detection
Find parcels on a belt and read their placement.
Helmet detection
Confirm a helmet is on before the line runs.
Fire and smoke
Catch smoke early in a fixed camera view.
License plates
Locate plates at a gate for a reader to parse.
Posture checks
Detect the body positions that matter as their own classes.
Fruit classification
Sort produce by ripeness and grade.
Medical imaging
Classify captured images in a controlled setup.
Industrial automation
Give a controller a vision signal it can act on.
Robot vision
Let a robot see the part before it reaches for it.
Deployment options
Where this runs
The targets teams reach for on this kind of project.
Raspberry Pi
Python inference package
Installs with the device client and runs as a service
UNO Q
Linux runtime package
The on-device runner takes it from there, no glue code to write
TensorFlow Lite
TensorFlow Lite model with an inference library
Embed it in any runtime that can load TensorFlow Lite
Petal Edge builds for six targets in total, each with its own package format and constraints.
See all deployment options Why Petal Edge
One platform, start to finish
Data, labels, training, evaluation, post-processing, deployment and the fleet live in one place, with nothing to stitch together.
The whole lifecycle
Collect, label, train, evaluate, post-process, deploy and monitor without leaving the platform.
Models sized for edge hardware
NanoVision, Vision Pro and MobileNetV2 are built to run on devices, not on datacentre GPUs.
Six deployment targets
One trained model, packaged for the hardware you already run.
A fleet you can reach
Register devices, read their inference logs, and push new models over the air.
What’s included
- Dataset import from COCO, YOLO, Pascal VOC, Open Images and Edge Impulse JSON
- AI-assisted labeling with human review before anything is applied
- Confusion matrix, per-class metrics and model testing across versions
- Post-processing with NMS and IoU tracking
- Over-the-air model updates to registered devices
- Project export, so your data and models stay portable
Start with your own images
Create a project, bring your data, and take it through to a deployed model.