Issues when running E2E ObjectDetection sample in Azure

When deploying the unmodified E2E sample https://github.com/dotnet/machinelearning-samples/tree/v1.2/samples/csharp/end-to-end-apps/DeepLearning_ObjectDetection_Onnx into an Azure AppService some issues occur.
I suppose that one of the purposes of the sample is to show how PrectionEnginePool may be used in the Azure so these issues should be resolved.
Azure AppService is a supported scenario according to: https://docs.microsoft.com/en-us/dotnet/machine-learning/how-to-guides/serve-model-web-api-ml-net

1. **the AppService does not even start in the default 32-bit mode.** The result is:
502 - Web server received an invalid response while acting as a gateway or proxy server.
I think this is something related with https://github.com/aspnet/AspNetCore/issues/6097 but I am not sure.
Setting 64-bit mode resolves this and the app starts.

2. You must **change the relative path to the images** because the folder structure in the Azure is little bit different than when running locally.
In the Get method in the ObjectDetectionController.cs the line in Azure must read:
` string imageFileRelativePath = @"assets" + url;`
Similarly other relative paths.

3. **Now click on any of the images. DivisionByZero is raised.** This is described also in the https://github.com/dotnet/machinelearning/issues/4171

Distribuisci l’esempio DeepLearning_ObjectDetection_Onnx non modificato in un Azure AppService e inizia da ObjectDetectionController.cs, verificando i percorsi relativi delle immagini usati dal relativo metodo Get. Confronta il comportamento a 32 bit e a 64 bit, quindi fai clic sulle immagini di esempio; l’attività è completata quando l’app si avvia nella configurazione supportata, i percorsi vengono risolti e non si verifica alcun DivisionByZero, con issue 4171 come contesto correlato.

Scritto dal modello di indicizzazione a partire dal testo della issue.