Project File Format

The .oneai file is a JSON document that stores all configuration for a ONE AI project. It is created automatically when you set up a new project in ONE WARE Studio and updated as you modify settings through the UI.

File Structure

{
  "type": "imageDetection",
  "guid": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "version": "1.2.3.4",
  "data": { ... }
}

Field	Type	Description
type	string	Project type identifier. Currently "imageDetection"
guid	string	Unique project identifier (UUID)
version	string	ONE AI version that last saved the file
data	object	All project configuration (see below)

---

Data Object

The data object contains the full project state:

Core Settings

Key	Type	Description
annotationMode	string	"classes", "objects", or "segmentation"
capabilityMode	string	"basic" or "advanced"
fusionType	string	Image fusion mode: "single", "multi", "difference", "comparison", "stereo"
multiImageMode	string	Selected multi-image processing mode
advancedMultiImageSettings	object	Configuration for multi-image fusion

Labels and Groups

"labels": [
  {
    "name": "Defect",
    "id": 0,
    "groupId": 0,
    "color": 4278190335,
    "excludeFromTraining": false
  }
],
"groups": [
  { "name": "Default", "id": 0 }
]

Field	Type	Description
name	string	Display name
id	int	Unique numeric identifier
groupId	int	Parent group ID (labels only)
color	uint	ARGB color as unsigned 32-bit integer
excludeFromTraining	bool	Whether to exclude this label from training

---

Hardware Settings

Defines the target deployment hardware. ONE AI optimizes model architecture for these constraints.

Key	Type	Default	Description
hardwareType	string	"FPGA"	Target hardware: FPGA, FPGA SoC, CPU, GPU, TPU, MCU, Server, ASIC
prioritizeSpeedOptimization	bool	false	Prioritize inference speed over accuracy
computeCapability	number	1	Available compute capacity
computeCapabilityUnit	string	"TOPS"	Unit: TOPS, MOPS, KOPS
dspBlocks	number	40	Number of 8-bit multipliers (FPGA DSP blocks)
dspGroups	number	1	8-bit multipliers with sum per DSP block
prioritizeMemoryOptimization	bool	false	Prioritize memory efficiency
memoryLimit	number	1	Available memory
memoryLimitUnit	string	"GB"	Unit: KB, MB, GB
optimizeForParallelExecution	bool	false	Enable parallel execution optimization
quantizedCalculations	bool	false	Use quantized arithmetic
bitsPerValue	number	8	Bit width for quantized values (2–32)
fpgaClockSpeed	number	50	FPGA clock speed in MHz
maximumMemoryUsage	number	25	Maximum memory usage (0–100%)
maximumDspUsage	number	25	Maximum multiplier usage (0–100%)

---

Prefilters

Prefilters are stored as arrays corresponding to pipeline stages:

"preFiltersBegin": [...],
"preFiltersBeforeAugmentation": [...],
"preFiltersAfterAugmentation": [...],
"preFiltersEnd": [...]

Each filter entry:

{
  "id": "initialResize",
  "isEnabled": true,
  "settings": { ... }
}

Field	Type	Description
id	string	Filter type identifier
isEnabled	bool	Whether the filter is active
settings	object	Filter-specific parameters

See Prefilters for available filter types and their parameters.

---

Augmentations

Augmentations are organized into three arrays:

"augmentationsBegin": [...],
"augmentationsStatic": [...],
"augmentationsDynamic": [...]

Each augmentation entry uses the same schema as prefilters:

{
  "id": "move",
  "isEnabled": true,
  "settings": { ... }
}

Default static augmentations: mosaic, move, rotate, flip, resize. Default dynamic augmentation: color.

See Augmentations for available augmentation types and their parameters.

---

Model Output Settings

Controls what the model predicts. Available keys depend on annotationMode:

Key	Modes	Type	Default	Description
classificationType	classes	string	"allIndividualClasses"	allIndividualClasses, oneClassPerImage, atLeastOneClass, regression
predictionType	objects	string	"sizePositionClass"	sizePositionClass, positionClass, allPresentClasses, largestArea, mostObjects, atLeastOneObject
segmentationType	segmentation	string	"oneClassPerPixel"	oneClassPerPixel, sizePositionClass, positionClass, allPresentClasses, largestArea, mostObjects, atLeastOneObject
sizePredictionEffort	objects	number	25	Effort allocated to size prediction (0–100%)
positionPredictionResolution	objects	number	—	Resolution for coordinate prediction
precisionRecallPrioritization	all	number	—	Balance between precision and recall

---

Model Input Settings (Advanced)

Fine-grained control over model architecture. These are automatically derived from Basic Mode settings unless overridden.

Key	Type	Default	Description
surroundingSizeMode	string	"relativeToObject"	Context sizing: relativeToObject or relativeToImage
minRelativeSurroundingSize	number	100	Minimum surrounding context (%)
maxRelativeSurroundingSize	number	100	Maximum surrounding context (%)
estimatedMinWidth	number	10	Estimated minimum object width (% of image)
estimatedMinHeight	number	10	Estimated minimum object height (% of image)
estimatedAvgWidth	number	50	Estimated average object width (%)
estimatedAvgHeight	number	50	Estimated average object height (%)
estimatedMaxWidth	number	90	Estimated maximum object width (%)
estimatedMaxHeight	number	90	Estimated maximum object height (%)
detectComplexity	number	50	Feature detection complexity (0–100%)
sameClassDifference	number	50	Intra-class variance (0–100%)
backgroundDifference	number	50	Object-to-background contrast (0–100%)
maxFeatures	number	10	Maximum features for classification
avgFeatures	number	2	Average features for classification

---

Model Input Settings (Basic Mode)

Simplified settings that auto-configure advanced model input parameters:

Key	Type	Default	Description
estimatedSize	string	"Small,Medium"	Object size categories: Tiny, Small, Medium, Big (comma-separated)
numberOfFeatures	string	"FewFeatures"	AlwaysOne, FewFeatures, ManyFeatures
typeOfEnvironment	string	"Controlled"	Controlled, Limited, Natural
typeOfFeatures	string	"Limited"	Similar, Limited, Open

---

Validation and Test Settings

Key	Type	Default	Description
useValidationSplit	bool	true	Enable automatic validation split
validationSplit	number	20	Percentage of training data used for validation (0–100)
testImagePercentage	number	0	Percentage of training images used for testing
validationImagePercentage	number	100	Percentage of validation images used for testing
validationImageSplitPercentage	number	0	Additional validation/test split

---

Auto-Label Settings

Key	Type	Description
selectedAutoLabelModels	array	Models used for auto-labeling: [{ "name": "...", "minConfidence": 0.5 }]
autoLabelMergeThreshold	number	Overlap threshold for merging auto-label predictions
selectedRunModel	string	Name of the selected inference model

---

Project Directory Structure

A .oneai project file sits alongside its associated data:

MyProject/
├── MyProject.oneai        ← project configuration
├── Dataset/               ← training images and annotations
├── Models/                ← trained model files
└── Export/                ← exported models (ONNX, VHDL, TFLite)

Need Help? We're Here for You!

Christopher from our development team is ready to help with any questions about ONE AI usage, troubleshooting, or optimization. Don't hesitate to reach out!

Our Support Email:support@one-ware.com