""" SDK estimator handler"""
from typing import List
import msgpack
from sanic.response import HTTPResponse
from app.api_sdk_adaptors.base import LoopEstimationsAlwaysOn, LoopEstimationsExtended
from app.api_sdk_adaptors.orientation import handleImageOrientation
from app.api_sdk_adaptors.sdk_adaptor import APISDKAdaptor
from app.handlers.base_handler import BaseHandlerWithMultipart
from app.handlers.custom_query_getters import (
int0180Getter,
livenessStatesValidator,
maskStatesValidator,
multifacePolicyGetter,
)
from classes.image_meta import InputImageData
from classes.multipart_processing import SDKMultipartProcessor
from classes.schemas.sdk import SDKInputEstimationsModel
from configs.config import LIVENESS_V2_QUALITY_THRESHOLD
from crutches_on_wheels.errors.errors import ErrorInfo
from crutches_on_wheels.errors.exception import VLException
from crutches_on_wheels.monitoring.points import monitorTime
from crutches_on_wheels.web.query_getters import boolFrom01Getter, float01Getter, int01Getter
from sdk.sdk_loop.enums import LoopEstimations, MultifacePolicy
from sdk.sdk_loop.models.image import ImageType
from sdk.sdk_loop.task import HandlersTask
from sdk.sdk_loop.tasks.filters import FaceDetectionFilters, Filters
from sdk.sdk_loop.tasks.task import LivenessV1Params, TaskEstimationParams, TaskParams
[docs]class SDKHandler(BaseHandlerWithMultipart):
"""
SDK estimator handler
Resource: "/{api_version}/sdk"
"""
[docs] async def getDataFromMultipart(self, imageType: ImageType = ImageType.IMAGE) -> List[InputImageData]:
"""Description see :func:`~BaseHandlerWithMultipart.getDataFromMultipart`."""
dataFromRequest = await SDKMultipartProcessor().getData(self.request)
estimationDataFromMultiPart = self._getDataFromMultipart(dataFromRequest.images, imageType)
return estimationDataFromMultiPart
[docs] async def post(self) -> HTTPResponse:
"""
SDK estimations handler. See `spec sdk`_.
.. _`spec sdk`:
_static/api.html#operation/sdk
Returns:
Response with estimations
"""
targets = set()
if self.getQueryParam("detect_face", int01Getter, default=0):
targets.add(LoopEstimations.faceDetection)
if self.getQueryParam("estimate_landmarks5", int01Getter, default=0):
targets.add(LoopEstimations.faceDetection)
targets.add(LoopEstimationsAlwaysOn.faceLandmarks5)
if self.getQueryParam("estimate_landmarks68", int01Getter, default=0):
targets.add(LoopEstimations.faceLandmarks68)
if self.getQueryParam("estimate_liveness", int01Getter, default=0):
targets.add(LoopEstimations.livenessV1)
if self.getQueryParam("estimate_head_pose", int01Getter, default=0):
targets.add(LoopEstimations.headPose)
if self.getQueryParam("estimate_gaze", int01Getter, default=0):
targets.add(LoopEstimations.gaze)
if self.getQueryParam("estimate_eyes_attributes", int01Getter, default=0):
targets.add(LoopEstimations.eyes)
if self.getQueryParam("estimate_mouth_attributes", int01Getter, default=0):
targets.add(LoopEstimations.mouthAttributes)
if self.getQueryParam("estimate_emotions", int01Getter, default=0):
targets.add(LoopEstimations.emotions)
if self.getQueryParam("estimate_mask", int01Getter, default=0):
targets.add(LoopEstimations.mask)
if self.getQueryParam("estimate_glasses", int01Getter, default=0):
targets.add(LoopEstimations.glasses)
if self.getQueryParam("estimate_face_warp", int01Getter, default=0):
targets.add(LoopEstimations.faceWarp)
if self.getQueryParam("estimate_quality", int01Getter, default=0):
targets.add(LoopEstimations.faceWarpQuality)
if self.getQueryParam("estimate_basic_attributes", int01Getter, default=0):
targets.add(LoopEstimations.basicAttributes)
if self.getQueryParam("estimate_face_descriptor", int01Getter, default=0):
targets.add(LoopEstimations.faceDescriptor)
if self.getQueryParam("detect_body", int01Getter, default=0):
targets.add(LoopEstimations.bodyDetection)
if self.getQueryParam("estimate_body_warp", int01Getter, default=0):
targets.add(LoopEstimations.bodyWarp)
if self.getQueryParam("estimate_body_descriptor", int01Getter, default=0):
targets.add(LoopEstimations.bodyDescriptor)
if self.getQueryParam("estimate_upper_body", int01Getter, default=0):
targets.add(LoopEstimations.bodyAttributes)
targets.add(LoopEstimationsExtended.upperBody)
if self.getQueryParam("estimate_body_basic_attributes", int01Getter, default=0):
targets.add(LoopEstimations.bodyAttributes)
targets.add(LoopEstimationsExtended.bodyBasicAttributes)
if self.getQueryParam("estimate_accessories", int01Getter, default=0):
targets.add(LoopEstimations.bodyAttributes)
targets.add(LoopEstimationsExtended.accessories)
pitchThreshold = self.getQueryParam("pitch_threshold", int0180Getter)
rollThreshold = self.getQueryParam("roll_threshold", int0180Getter)
yawThreshold = self.getQueryParam("yaw_threshold", int0180Getter)
scoreThreshold = self.getQueryParam("score_threshold", float01Getter)
maskStates = self.getQueryParam("mask_states", maskStatesValidator)
livenessStates = self.getQueryParam("liveness_states", livenessStatesValidator)
faceFilters = FaceDetectionFilters(
yawThreshold=yawThreshold,
pitchThreshold=pitchThreshold,
rollThreshold=rollThreshold,
gcThreshold=scoreThreshold,
maskStates=maskStates,
livenessStates=livenessStates,
)
params = TaskParams(
targets=targets,
filters=Filters(faceDetection=faceFilters),
estimatorsParams=TaskEstimationParams(
livenessv1=LivenessV1Params(qualityThreshold=LIVENESS_V2_QUALITY_THRESHOLD),
faceDescriptorVersion=self.config.defaultFaceDescriptorVersion,
bodyDescriptorVersion=self.config.defaultHumanDescriptorVersion,
),
multifacePolicy=self.getQueryParam(
"multiface_policy", multifacePolicyGetter, default=MultifacePolicy.allowed
),
useExifInfo=self.getQueryParam("use_exif_info", boolFrom01Getter, default=True),
autoRotation=self.config.useAutoRotation,
aggregate=self.getQueryParam("aggregate_attributes", int01Getter, default=0),
)
imageType = self.getQueryParam("image_type", lambda x: ImageType(int(x)), default=ImageType.IMAGE)
with monitorTime(self.request.dataForMonitoring, "load_images_for_processing_time"):
inputData = await self.getInputEstimationData(
self.request, imageType=imageType, validationModel=SDKInputEstimationsModel
)
self.assertLicence(
liveness=(LoopEstimations.livenessV1 in targets or faceFilters.livenessFilter),
bodyAttributes=(LoopEstimations.bodyAttributes in targets),
)
task = HandlersTask(data=[metaImage.image for metaImage in inputData], params=params)
await task.execute()
if task.result.error:
raise VLException(ErrorInfo.fromDict(task.result.error.asDict()), 400, isCriticalError=False)
if self.config.useAutoRotation:
handleImageOrientation(task.result.images)
encodeToBase64 = True
if (responseContentType := self.getResponseContentType()) == "application/msgpack":
encodeToBase64 = False
sdkAdaptor = APISDKAdaptor(
estimationTargets=targets, aggregationEnabled=params.aggregate, encodeBytesToBase64=encodeToBase64
)
result, monitoringData = await sdkAdaptor.buildResult(
task.result, meta=[metaImage.meta for metaImage in inputData]
)
self.countLivenessEstimationsPerformed(monitoringData.sdkUsages.livenessEstimator)
self.handleMonitoringData(monitoringData)
if responseContentType == "application/msgpack":
body = msgpack.packb(result, use_bin_type=True)
return self.success(200, body=body, contentType="application/msgpack")
return self.success(200, outputJson=result)
