vart Namespace Reference

VART (Vitis AI Runtime) public API namespace. More...

Classes
class	DeviceImplBase
	Abstract interface for device implementation used by Device. More...
class	InferResultImplBase
	Abstract base class for inference result implementation used by InferResult. More...
class	MemoryImplBase
	Abstract interface for memory implementation used by Memory. More...
class	MetaConvertImplBase
	Abstract interface for metaconvert implementation used by MetaConvert. More...
class	OverlayImplBase
	Abstract interface for overlay implementation used by Overlay. More...
class	PLKernelImplBase
	Abstract interface for PLKernel implementation used by PLKernel. More...
class	PostProcessImplBase
	Abstract interface for post-processing implementation used by PostProcess. More...
class	PreProcessImplBase
	Abstract interface for pre-processing implementation used by PreProcess. More...
class	VideoFrameImplBase
	Abstract base class for video frame implementation used by VideoFrame. More...
class	Device
	Manages the hardware context and loading of xclbin onto the device. Any module utilizing hardware acceleration requires a Device instance. Instances are not constructed directly; use the static factory get_device_hdl(), which caches results per (dev_idx, xclbin_loc). More...
class	InferResult
	This module is used to represent inference results. Built-in supported types are ROOT, CLASSIFICATION, DETECTION and SEGMENTATION. Custom types can be integrated by deriving from InferResultImplBase and passing the shared pointer to the pimpl constructor; the CUSTOM_RESULT_* enum values are NOT accepted by the type-enum constructor below. More...
struct	InferResScaleInfo
	Geometry consumed by InferResult::transform() to map results to the original input frame. More...
struct	InferResultData
	Base structure for inference result data. More...
struct	ClassificationResData
	Classification infer result. More...
struct	DetectionResData
	Detection infer result. More...
struct	SegmentationResData
	Segmentation infer result. More...
class	Logger
	The logger module provides the logging support for VART modules. Supports logging to console, file, syslog. More...
struct	log_context
	Context of a particular module instance. More...
struct	LogBackend
class	Memory
	This module is responsible for allocating and managing memory on the device. More...
class	MetaConvert
	This module facilitates the conversion of Infer metadata into a format compatible with the overlay module. Metaconvert also accepts configuration parameters as JSON string, which provide further flexibility on configuring overlay information such as line thickness, font size, font type ., etc. Please check API documentation for more information. Additionally, if users have a custom meta data then they can integrate customized functions to convert them into a format suitable for processing by the overlay module by overriding base class. More...
class	Overlay
	This module facilitates the overlay of annotations onto the video frame, currently overlay utilizes OpenCV library to draw on frames, which is software based. Overlay supports drawing of bounding boxes, text, lines, arrows, circles and polygons on frames. Application can also incorporate custom implementation using base class. More...
struct	OverlayCoordinates
	Structure representing Coordinate information. More...
struct	OverlayColorData
	Structure representing Color information. More...
struct	OverlayRectParams
	Structure representing information to draw rectangle on frame. More...
struct	OverlayFontData
	Structure representing Font information. More...
struct	OverlayTextParams
	Structure representing Text information. More...
struct	OverlayLineParams
	Structure representing Line information. More...
struct	OverlayArrowParams
	Structure representing Arrow information. More...
struct	OverlayCircleParams
	Structure representing Circle Information. More...
struct	OverlayPolygonParams
	Structure representing Polygon information. More...
struct	OverlayMaskParams
	Structure representing Mask information. More...
struct	OverlayShapeInfo
	Structure representing overlay shape information. More...
class	PLKernel
	PLKernel class for managing PL kernel execution. More...
class	ArgumentInfo
	Contains information related to an argument for the PL kernel. More...
class	PostProcess
	Interface class for post-processing inference results. More...
struct	TensorInfo
	Tensor config information. More...
class	PreProcess
	The preprocessing module handles data preparation tasks such as normalization, scaling, and video format conversion. This module supports software-based pre-processing as well as hardware-accelerated pre-processing for optimized performance. It ensures that input data is appropriately formatted for inference. Applications can also supply a custom pre-processing implementation via the PreProcessImplBase constructor. More...
struct	PreProcessInfo
	Structure storing pre-processing parameters. More...
struct	RegionOfInterest
	Defines a rectangular region of interest within a frame. More...
struct	PreProcessTypeMeta
	Preprocess-owned geometry recorded after add_frame. More...
struct	PreProcessOpRes
	Per-op outputs returned by PreProcess::process(). More...
struct	PreProcessOp
	Describes a single pre-processing operation, pairing an input frame and its ROI with an output frame and its ROI. More...
class	VideoFrame
	This module simplifies the management of video frame memory and provides APIs for mapping and unmapping frame data to user space to read and write. The VideoFrame class offers flexibility for applications to encapsulate their own memory into the VideoFrame class. In such instances, the application bears the responsibility for deallocating the frame memory. More...
struct	VideoAlignment
	Contains video alignment information. More...
struct	VideoInfo
	Contains information related to a video frame. More...
struct	VideoPlaneInfo
	Structure containing information specific to a video frame plane. More...
struct	VideoFrameMapInfo
	Structure encapsulating details of a video frame after a map operation. More...

Typedefs
using	LogAcceptFn = bool(*)(LogLevel level, const char *basename, int32_t instance_id, void *user_data)
using	LogDeliverFn = void(*)(LogLevel level, const char *basename, int32_t instance_id, const char *file, const char *func, uint32_t line, const char *message, void *user_data)

Enumerations
enum class	TraversalOrder { PREORDER , INORDER , POSTORDER }
enum class	InferResultType {   ROOT , CLASSIFICATION , DETECTION , SEGMENTATION ,   CUSTOM_RESULT_1 = 100 , CUSTOM_RESULT_2 , CUSTOM_RESULT_3 }
enum class	SegmentationType { SEMANTIC = 0 , MEDICAL , SEG3D }
enum class	LogLevel {   LOG_LEVEL_NONE = 0 , LOG_LEVEL_ERROR , LOG_LEVEL_WARNING , LOG_LEVEL_FIXME ,   LOG_LEVEL_INFO , LOG_LEVEL_DEBUG }
enum	LogType { NO_LOG = 0 , CONSOLE , FILE , SYSLOG }
enum class	DataMapFlags { NONE = 0 , READ = 1 << 0 , WRITE = 1 << 1 }
enum class	MemoryImplType { XRT , NON_CMA }
enum class	OverlayImplType { OPENCV }
enum	OverlayArrowDirection { ARROW_DIRECTION_START , ARROW_DIRECTION_END , ARROW_DIRECTION_BOTH_ENDS }
	Enum representing arrow direction information. More...
enum class	OverlayMaskType { MASK_TYPE_LABELMAP , MASK_TYPE_COLORMAP }
	Enum representing mask type information. More...
enum class	PLKernelImplType { PL_KERNEL_XRT }
enum class	TensorDataType {   INT8 , BF16 , FP16 , FLOAT32 ,   UNKNOWN }
	Specifies the data type of a tensor. More...
enum class	TensorDataDirection { INPUT , OUTPUT }
	Specifies the direction of a tensor (input or output). More...
enum class	PostProcessType {   RESNET50 , YOLOV2 , SSDRESNET34 , SOFTMAX ,   ARGMAX , TOPK , NMS , THRESHOLD ,   LABEL_MAPPING , NORMALIZATION , ANCHOR_ADJUSTMENT , CALIBRATION_TEMPERATURE ,   CALIBRATION_PLATT , BIAS_CORRECTION , OUTLIER_DETECTION , UNCERTAINTY_ESTIMATION ,   DISTANCE_IOU_NMS , SOFT_NMS , CLASSWISE_NMS , OBJECT_COUNT ,   SOFTMAXSEG , SIGMOIDSEG , ARGMAXSEG }
enum class	PreProcessImplType { IMAGE_PROCESSING_HLS , IMAGE_PROCESSING_SW }
	Selects the pre-processing backend implementation to use. More...
enum class	PreProcessType { DEFAULT , LETTERBOX , PANSCAN }
	Selects the scaling/resize policy applied during pre-processing. More...
enum class	VideoFrameImplType { XRT , NON_CMA }
	Video Frame implementation type. More...
enum class	VideoFormat {   UNKNOWN = 0 , Y_UV8_420 , RGBx , r210 ,   Y410 , BGRx , BGRA , RGBA ,   YUY2 , NV16 , RGB , v308 ,   BGR , I422_10LE , NV12_10LE32 , GRAY8 ,   GRAY10_LE32 , I420 , RGBP , BGR_FLOAT ,   RGB_FLOAT , RGBP_FLOAT , GRAY32_FLOAT , RGBx_BF16 ,   BGRx_BF16 , GRAY_BF16 , RGBx_FP16 , BGRx_FP16 ,   GRAY_FP16 , RGB_FP16 , BGR_FP16 , RGB_BF16 ,   BGR_BF16 , RGBP_FP16 , RGBP_BF16 , BGRP ,   BGRP_FP16 , BGRP_BF16 , BGRP_FLOAT }
	Represents video color formats supported by VART core APIs. More...

Detailed Description

VART (Vitis AI Runtime) public API namespace.

Typedef Documentation

LogAcceptFn

using vart::LogAcceptFn = typedef bool (*)(LogLevel level, const char *basename, int32_t instance_id, void *user_data)

LogDeliverFn

using vart::LogDeliverFn = typedef void (*)(LogLevel level, const char *basename, int32_t instance_id, const char *file, const char *func, uint32_t line, const char *message, void *user_data)

Enumeration Type Documentation

DataMapFlags

enum vart::DataMapFlags

strong

• Map flags.

These values are bit flags and may be combined with bitwise OR (e.g. READ|WRITE) when calling map() to request read/write mappings.

Enumerator
NONE	Default value; no specific access mode requested.
READ	Map data in read mode.
WRITE	Map data in write mode.

InferResultType

enum vart::InferResultType

strong

• ML Network type

Enumerator
ROOT	Root node of the tree structure.
CLASSIFICATION	Classification network.
DETECTION	Detection network.
SEGMENTATION	Segmentation network (per-pixel class maps).
CUSTOM_RESULT_1	Defined for user specific result type.
CUSTOM_RESULT_2	Defined for user specific result type.
CUSTOM_RESULT_3	Defined for user specific result type.

LogLevel

enum vart::LogLevel

strong

• Log levels supported.

Levels are ordered by verbosity (lowest underlying integer value is the quietest). When configured via the VART_CORE_DEBUG_ENV environment variable, levels are expected as their numeric values (0..5) — symbolic names are not parsed.

Enumerator
LOG_LEVEL_NONE	No logging output.
LOG_LEVEL_ERROR	Print ERROR logs.
LOG_LEVEL_WARNING	Print WARNING and ERROR logs.
LOG_LEVEL_FIXME	Print any TODO/FIXME logs.
LOG_LEVEL_INFO	Print INFO, WARNING & ERROR logs.
LOG_LEVEL_DEBUG	Print DEBUG, INFO, WARNING & ERROR logs.

LogType

enum vart::LogType

• Log types(sink) supported

Enumerator
NO_LOG	Do not log to any sink.
CONSOLE	Logs to console.
FILE	Logs to file specified with VART_LOG_FILE_PATH environment variable.
SYSLOG	Logs using syslog mechanism.

MemoryImplType

enum vart::MemoryImplType

strong

• Memory Implementation type.

Enumerator
XRT	XRT allocation type.
NON_CMA	Non CMA allocation type.

OverlayArrowDirection

enum vart::OverlayArrowDirection

Enum representing arrow direction information.

Enumerator
ARROW_DIRECTION_START	Arrow head drawn at the start point.
ARROW_DIRECTION_END	Arrow head drawn at the end point.
ARROW_DIRECTION_BOTH_ENDS	Arrow heads drawn at both ends.

OverlayImplType

enum vart::OverlayImplType

strong

• Overlay Implementation type

Enumerator
OPENCV	OpenCV-based implementation.

OverlayMaskType

enum vart::OverlayMaskType

strong

Enum representing mask type information.

Enumerator
MASK_TYPE_LABELMAP	Mask type label map.
MASK_TYPE_COLORMAP	Mask type color map.

PLKernelImplType

enum vart::PLKernelImplType

strong

• PL Kernel Implementation type

Enumerator
PL_KERNEL_XRT	PL kernel implementation using XRT APIs.

PostProcessType

enum vart::PostProcessType

strong

• post-processing algorithm types

Enumerator
RESNET50	Resnet50 model.
YOLOV2	Yolov2 model.
SSDRESNET34	ssdresnet34 model
SOFTMAX	Softmax activation function.
ARGMAX	Argmax function to find index of maximum value.
TOPK	TopK function to find k largest elements.
NMS	Non-Maximum Suppression.
THRESHOLD	Thresholding function.
LABEL_MAPPING	Map class indices to labels.
NORMALIZATION	Normalize values.
ANCHOR_ADJUSTMENT	Adjust bounding boxes based on anchors.
CALIBRATION_TEMPERATURE	Temperature scaling calibration.
CALIBRATION_PLATT	Platt scaling calibration.
BIAS_CORRECTION	Bias correction.
OUTLIER_DETECTION	Outlier detection.
UNCERTAINTY_ESTIMATION	Uncertainty estimation.
DISTANCE_IOU_NMS	Distance intersection over union NMS.
SOFT_NMS	Soft NMS.
CLASSWISE_NMS	Class-wise NMS.
OBJECT_COUNT	Object counting.
SOFTMAXSEG	Softmax for segmentation.
SIGMOIDSEG	Sigmoid for segmentation.
ARGMAXSEG	Argmax for segmentation.

PreProcessImplType

enum vart::PreProcessImplType

strong

Selects the pre-processing backend implementation to use.

Enumerator
IMAGE_PROCESSING_HLS	Implementation using the Image Processing HLS kernel.
IMAGE_PROCESSING_SW	Implementation using the Image Processing software kernel.

PreProcessType

enum vart::PreProcessType

strong

Selects the scaling/resize policy applied during pre-processing.

• DEFAULT : Direct stretch-to-fit resize. maintain_aspect_ratio and symmetric_padding in PreProcessInfo are only evaluated for this type and are ignored for PANSCAN and LETTERBOX.
• LETTERBOX : Resize while maintaining aspect ratio; the output is padded with symmetric_padding=true (centered, standard convention).
• PANSCAN : Center-crop the input ROI to the output aspect ratio, then resize the cropped region to fill the output without padding.

Enumerator
DEFAULT	Stretch resize; maintain_aspect_ratio/symmetric_padding apply only here.
LETTERBOX	Maintain aspect ratio with centered symmetric padding.
PANSCAN	Center-crop to output aspect ratio, then resize to fill output.

SegmentationType

enum vart::SegmentationType

strong

• Segmentation type

Enumerator
SEMANTIC	Semantic segmentation.
MEDICAL	Medical segmentation.
SEG3D	3D segmentation

TensorDataDirection

enum vart::TensorDataDirection

strong

Specifies the direction of a tensor (input or output).

Enumerator
INPUT	Represents an input tensor.
OUTPUT	Represents an output tensor.

TensorDataType

enum vart::TensorDataType

strong

Specifies the data type of a tensor.

Enumerator
INT8	Represents an 8-bit integer data type.
BF16	Represents a brain floating-point 16-bit data type.
FP16	Represents a standard 16-bit floating-point data type.
FLOAT32	Represents a 32-bit floating-point data type.
UNKNOWN	Sentinel value for an unrecognized or unset data type; used to detect misconfigured tensors and reject them early (e.g. in set_config).

TraversalOrder

enum vart::TraversalOrder

strong

• Tree traversal order for infer result tree

Enumerator
PREORDER	Traversal order where the parent node is processed before its children.
INORDER	Traversal order where the parent node is processed between its children.
POSTORDER	Traversal order where the parent node is processed after its children.

VideoFormat

enum vart::VideoFormat

strong

Represents video color formats supported by VART core APIs.

Enumerator
UNKNOWN	Unknown color format.
Y_UV8_420	Planar 4:2:0 YUV with interleaved UV plane.
RGBx	Packed RGB, 4 bytes per pixel.
r210	Packed 4:4:4 RGB, 10 bits per channel.
Y410	Packed 4:4:4 YUV, 10 bits per channel.
BGRx	Packed BGR, 4 bytes per pixel.
BGRA	Packed BGR with alpha channel last.
RGBA	Packed RGB with alpha channel last.
YUY2	Packed 4:2:2 YUV (Y0-U0-Y1-V0, Y2-U2-Y3-V2...)
NV16	Planar 4:2:2 YUV with interleaved UV plane.
RGB	RGB packed into 24 bits without padding.
v308	Packed 4:4:4 YUV.
BGR	BGR packed into 24 bits without padding.
I422_10LE	Planar 4:2:2 YUV, 10 bits per channel.
NV12_10LE32	10-bit variant of NV12, packed into 32bit words (MSB 2 bits padding)
GRAY8	8-bit grayscale
GRAY10_LE32	10-bit grayscale, packed into 32bit words (2 bits padding)
I420	Planar 4:2:0 YUV.
RGBP	Planar RGB.
BGR_FLOAT	BGR float, 96 bits without padding (32 bits per component)
RGB_FLOAT	RGB float, 96 bits without padding (32 bits per component)
RGBP_FLOAT	RGB float, planar (32 bits per component)
GRAY32_FLOAT	Grayscale with 32-bit float data.
RGBx_BF16	Packed RGB (Brain Float), 8 bytes per pixel.
BGRx_BF16	Packed BGR (Brain Float), 8 bytes per pixel.
GRAY_BF16	Grayscale with 16-bit Brain Float data.
RGBx_FP16	Packed RGB (Half Precision Float), 8 bytes per pixel.
BGRx_FP16	Packed BGR (Half Precision Float), 8 bytes per pixel.
GRAY_FP16	Grayscale with 16-bit Half Precision Float data.
RGB_FP16	RGB packed (Half Precision Float), 6 bytes per pixel.
BGR_FP16	BGR packed (Half Precision Float), 6 bytes per pixel.
RGB_BF16	RGB packed (Brain Float), 6 bytes per pixel.
BGR_BF16	BGR packed (Brain Float), 6 bytes per pixel.
RGBP_FP16	RGB planar (Half Precision Float), 16 bits per component.
RGBP_BF16	RGB planar (Brain Float), 16 bits per component.
BGRP	Planar BGR, 8 bits per component.
BGRP_FP16	BGR planar (Half Precision Float), 16 bits per component.
BGRP_BF16	BGR planar (Brain Float), 16 bits per component.
BGRP_FLOAT	BGR float, planar (32 bits per component)

VideoFrameImplType

enum vart::VideoFrameImplType

strong

Video Frame implementation type.

Enumerator
XRT	Allocate memory using Xilinx Run Time (XRT) framework. Allocates physically contiguous memory.
NON_CMA	Non-contiguous memory type.