We're running the container preview in AKS and our containers are filled with the following exception:

fail: Microsoft.CloudAI.Containers.Diagnostics.ExceptionMiddleware[0]

  Could not load type 'System.Net.Http.HttpStatusCodeExtensions' from assembly 'Microsoft.CloudAI.Containers.Common, Version=1.1.1201.1, Culture=neutral, PublicKeyToken=31bf3856ad364e35'. SubscriptionId='' RequestId='d1845cb5-99d0-46e3-b3e0-5ebf81ac51d9' Timestamp=''

System.TypeLoadException: Could not load type 'System.Net.Http.HttpStatusCodeExtensions' from assembly 'Microsoft.CloudAI.Containers.Common, Version=1.1.1201.1, Culture=neutral, PublicKeyToken=31bf3856ad364e35'.
at Microsoft.CloudAI.Containers.Controllers.VisionControllerBase.CreateErrorResponse(HttpStatusCode statusCode, String code, String message)
at Microsoft.CloudAI.Containers.OneOcr.OcrControllerBase.GetTextOperationResultWorker(Guid id) in /source/src/OneOcr.Common/Controllers/OcrControllerBase.cs:line 92
at Microsoft.CloudAI.Containers.OneOcr.OcrController.GetTextOperationResult(Guid id) in /source/src/OneOcr.2.0/Controllers/OcrController.cs:line 108
at Microsoft.AspNetCore.Mvc.Internal.ActionMethodExecutor.TaskOfIActionResultExecutor.Execute(IActionResultTypeMapper mapper, ObjectMethodExecutor executor, Object controller, Object[] arguments)
at Microsoft.AspNetCore.Mvc.Internal.ControllerActionInvoker.InvokeActionMethodAsync()
at Microsoft.AspNetCore.Mvc.Internal.ControllerActionInvoker.InvokeNextActionFilterAsync()
at Microsoft.AspNetCore.Mvc.Internal.ControllerActionInvoker.Rethrow(ActionExecutedContext context)
at Microsoft.AspNetCore.Mvc.Internal.ControllerActionInvoker.Next(State& next, Scope& scope, Object& state, Boolean& isCompleted)
at Microsoft.AspNetCore.Mvc.Internal.ControllerActionInvoker.InvokeInnerFilterAsync()
at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.InvokeNextResourceFilter()
at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.Rethrow(ResourceExecutedContext context)
at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.Next(State& next, Scope& scope, Object& state, Boolean& isCompleted)
at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.InvokeFilterPipelineAsync()
at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.InvokeAsync()
at Microsoft.AspNetCore.Builder.RouterMiddleware.Invoke(HttpContext httpContext)
at Microsoft.AspNetCore.StaticFiles.StaticFileMiddleware.Invoke(HttpContext context)
at Swashbuckle.AspNetCore.SwaggerUI.SwaggerUIIndexMiddleware.Invoke(HttpContext httpContext)
at Swashbuckle.AspNetCore.Swagger.SwaggerMiddleware.Invoke(HttpContext httpContext)
at Microsoft.CloudAI.Containers.Diagnostics.RequestCorrelationMiddleware.InvokeRequestCorrelationMiddleware(HttpContext httpContext)
at Microsoft.CloudAI.Containers.Diagnostics.RequestCorrelationMiddleware.Invoke(HttpContext httpContext)
at Microsoft.CloudAI.Containers.Diagnostics.ExceptionMiddleware.Invoke(HttpContext httpContext)

Is this expected? Is there a resolution?

When testing with Turkish text that also includes numerical strings, I have isolated the following anomaly:

Consider two versions of an otherwise identical image (for informative purposes, the image is a graphic depicting the total number of deaths in Turkey from COVID-19, on a particular date):

Turkish version:
TÜRKİYE’DE ÖLÜMLER 1.368

English version:
DEATHS IN TURKEY 1.368

When the image is OCR-processed with the text in English (i.e. with the words: “DEATHS IN TURKEY”), the numeric string is returned correctly as “1.368”.

However, when the image is OCR-processed with the text in Turkish (i.e. with the words: “TÜRKİYE’DE ÖLÜMLER”), the numeric string is recognized incorrectly, such that a Turkish lowercase dotless i ("ı", Unicode U+0131) is returned for the digit “1” in the numeric string, while the remaining part is returned correctly as “.368”.

In both images, the numerical value is in the exact same location. Backgrounds of both images are identical. And note that the thousands separator character used in Turkey is period (not comma, as used in the U.S.).

I feel recognition accuracy of numeric strings should not degrade when the primary language detected for the request is Turkish.

I used the following endpoints for testing, with identical results:
COMPUTER_VISION_ENDPOINT/vision/v1.0/ocr
COMPUTER_VISION_ENDPOINT/vision/v2.1/ocr

I have attached below both images, as well as the JSON output returned for both requests.

Following curl request may be used to test the images:

curl \
--header "ocp-apim-subscription-key: <azure.subscription.id>" \
--header "content-type: application/octet-stream" \
--data-binary "@<image.file.full.pathname>" \
"https://<cognitive.resource>.cognitiveservices.azure.com/vision/v2.1/ocr"

Face API allows has 2 different face detection model.
The main difference (at least for us) is that detection_01 provides face attributes, while detection_02 does not, but it detects small, side-view and blurry faces.

We would be keen to use AnalyzeImage with detection_02 - we don't need face attributes but need to find as many faces as possible along with other objects and image description.

It's possible to do it now by making separate calls to AnalyzeImage and Face.DetectFaces, but it's not very good cost wise.

Would it be possible to extend AnalyzeImage API to allow setting face detection model?

The difference between the face detection models can be found here https://docs.microsoft.com/en-us/azure/cognitive-services/face/face-api-how-to-topics/specify-detection-model#evaluate-different-models.

If a document consists of pages that contain font of a 'good' size and then some documents (think Terms and Conditions) that contains a lot of small text that can't be interpreted by Computer Vision then it gets excluded from the response.

The page numbers are returned - so this is useful. But in the case where a 'complicated' page is the last page, this would not be returned, so the number of pages you believe you have would be one less than the number of pages you actually have. Including (at the very least) a JSON segment with:
{ "page":# }
would ensure that all pages can be accurately recognised by a system taking advantage of this api.