I've create a custom vision project to recognise characters (A, B, C...).
What is interesting: if I upload an image of a character (in this case an "N") to the vision API portal it will tell me that it is 99.9% sure it is an "N":
If however I use the client libraries to predict the very same image, I'm getting 53% that it is a "W" and only 37% that it is an "N":
I double checked that the latest iteration is the published one
I double checked that I'm using the correct project ID
My endpoint is set to "https://westeurope.api.cognitive.microsoft.com" in the CustomVisionPredictionClient
The code to get the prediction on my client:
var client = new CustomVisionPredictionClient()
{
ApiKey = predictionKey,
Endpoint = endpoint
};
var result = await client.PredictImageAsync(Guid.Parse(projectId), imageStream).ConfigureAwait(false);
var prediction = result.Predictions.FirstOrDefault();
Where does this difference come from and how to fix because according to the tests I did by uploading images the results are close to 100% correct no matter which character image I upload?
UPDATE: I noticed that there was an update for the client libraries. They went from 0.12pre to 1.0stable. After the update the PredictImageAsync is gone and replaced with DetectImageAsync. This expected as an additional parameter a model name. I tried using the name of the iteration and after a while the method returns with an internal server error. So not sure what to try next.
The comment above pointed my into the right direction - thanks!
The new client library has got two methods ClassifyImage and DetectImage (and various variations of them) which replace the previously used ones including PredictImage which I was using with the preview version of the client library.
To classify an image (which is what I wanted to do) ClassifyImage should of course be used. The new code looks like this and delivers an almost 100% correct prediction:
var client = new CustomVisionPredictionClient()
{
ApiKey = predictionKey,
Endpoint = endpoint
};
var result = await client.ClassifyImageAsync(Guid.Parse(projectId), "Iteration12", imageStream).ConfigureAwait(false);
var prediction = result.Predictions.FirstOrDefault();
endpoint is the URL of the region the vision API is hosted in, in my case https://westeurope.api.cognitive.microsoft.com.
predictionKey is available on the CustomVision.AI site in your project, so is the projectId
The publishedName parameter is the name of the iteration to use (in my case "Iteration12"
Related
I'm using .net-core3.1 with Microsoft.Azure.CognitiveServices.Language.SpellCheck NuGet package. I've read through entire documentation around Bing/cognitive API but I still find it very confusing as there are multiple APIs doing the same thing.
I got the API key from Microsoft.BingSearch on portal.azure.com and I'm using the free subscription. My subscription should however be valid as I am already using their LUIS without problems. Azure links to https://learn.microsoft.com/en-us/bing/search-apis/bing-spell-check/quickstarts/rest/python for quick start but this does not work for me ("https://api.bing.microsoft.com/v7.0/SpellCheck" url gives me "NotFound" using the code below with my key).
code sample:
var x = new SpellCheckClient(new ApiKeyServiceClientCredentials("<API_KEY>"));
// endpoints I tried:
// x.Endpoint = "https://westeurope.api.bing.microsoft.com/v7.0/spellcheck";
// x.Endpoint = "https://cognitiveservices.azure.com/bing/v7.0";
// x.Endpoint = "https://api.bing.microsoft.com"; -- Not found
// x.Endpoint = "https://cognitiveservices.azure.com"; -- The requested name is valid, but no data of the requested type was found.
var y = await x.SpellCheckerWithHttpMessagesAsync("gona");
Using default endpoint gives me Unauthorized error code.
Anyone has any idea on how to use this API?
You are right, the endpoint seems to be wrong. As you can see in the documentation here, regarding this value:
Supported Cognitive Services endpoints (protocol and hostname, for
example: "https://westus.api.cognitive.microsoft.com",
"https://api.cognitive.microsoft.com").
So if you are using West Europe, it should be "https://westus.api.cognitive.microsoft.com"
You can also check your API key by directly testing the console here: https://westeurope.dev.cognitive.microsoft.com/docs/services/5f7d486e04d2430193e1ca8f760cd7ed/operations/57855119bca1df1c647bc358
Choose your resource region (the one selected during key creation on Azure portal)
Set your key value in "Ocp-Apim-Subscription-Key" field
Edit the "text" value in the query parameters
Run the request
I'm trying to publish my Custom Vision Iteration after I trained it. But I always get a "Bad Request" Error.
I'm trying with the following line of code:
trainingApi.PublishIteration(ProjectID, iteration.Id, "Model", predictionResourceId);
It should Publish my Iteration but I just get an error.
I re-checked all my ID's but everything looks fine. Has the model name be something specific (start with lower letter or something)?
Edit:
I tried it now with a POST Request in Postman but now I receive:
{
"code": "BadRequestInvalidPublishTarget",
"message": "Invalid prediction resource id"
}
But I re-checked my Prediction Resource ID and it's correct.
Edit 2:
I thin I put the wrong thing into predictionId in the POST request, I just put in a ID but I think it should have been the /subscriptions/... part like described by microsoft. The problem now is:
{
"code": "BadRequestInvalidPublishTarget",
"message": "Invalid prediction id, please pass a prediction resource id."
}
For me this means it doesn't receive a prediction resource id, but I'm lost so I'm out of ideas what the problem could be.
Edit 3:
I forgot to add my POST Request:
https://xxx.cognitiveservices.azure.com/customvision/v3.0/training/projects/xxx/
iterations/xxx/publish?publishName=Model&predictionId=/subscriptions/xxx/
resourceGroups/CustomVision/providers/Microsoft.CognitiveServices/accounts/xxx
Publish using POST request
If you want to ensure that you have the right syntax, you can check using Custom vision portal by doing the same steps.
For example when I try to publish an iteration of a project, I can see the following call in the console:
https://westeurope.api.cognitive.microsoft.com/customvision/v3.3/Training/projects/ID_OF_MY_PROJECT/iterations/ID_OF_MY_ITERATION/publish?predictionId=%2Fsubscriptions%2FID_OF_MY_AZURE_SUBSCRIPTION%2FresourceGroups%2FNAME_OF_MY_RESOURCE_GROUP%2Fproviders%2FMicrosoft.CognitiveServices%2Faccounts%2FNAME_OF_MY_CUSTOM_VISION_PREDICTION_RESOURCE&publishName=NAME_OF_MY_ITERATION
Demo:
So yes, the "publicationId" value is looking like the one you mentioned but you have to encode the value of this string.
So change this:
predictionId=/subscriptions/xxx/
resourceGroups/CustomVision/providers/Microsoft.CognitiveServices/accounts/xxx
to
predictionId=%2Fsubscriptions%2Fxxx%2F
resourceGroups%2FCustomVision%2Fproviders%2FMicrosoft.CognitiveServices%2Faccounts%2Fxxx
in your call.
And be careful to use a prediction resource, not a training one.
Publish using C#
Here is a demo using C# and the official Custom Vision package hosted on Nuget (here)
using System;
using Microsoft.Azure.CognitiveServices.Vision.CustomVision.Training;
namespace so65714960
{
class Program
{
private static CustomVisionTrainingClient _trainingClient;
static void Main(string[] args)
{
Console.WriteLine("Hello World!");
_trainingClient = new CustomVisionTrainingClient(new ApiKeyServiceClientCredentials("PUT_YOUR_TRAINING_KEY_HERE"));
// I'm specifying my endpoint here as I'm working on West Europe region
_trainingClient.Endpoint = "https://westeurope.api.cognitive.microsoft.com/";
var projectId = new Guid("4b...a5"); // Put your Project Id here
var iterationId = new Guid("9d...e"); // Put your iteration Id here
// Get iteration information
var targetIteration = _trainingClient.GetIteration(projectId, iterationId);
Console.WriteLine($"Iteration publish resource Id: '{targetIteration.OriginalPublishResourceId}'");
// If originalPublishResourceId is not null, it is already published
// For this demo purpose, we unpublish first to publish again after if it is already published
if (!string.IsNullOrWhiteSpace(targetIteration.OriginalPublishResourceId))
{
_trainingClient.UnpublishIteration(projectId, iterationId);
// Force status refresh
targetIteration = _trainingClient.GetIteration(projectId, iterationId);
Console.WriteLine($"Iteration publish resource Id after unpublish: '{targetIteration.OriginalPublishResourceId}'");
}
// Publish
var publicationResourceId = "/subscriptions/7c...e8/resourceGroups/Cognitive_Demo/providers/Microsoft.CognitiveServices/accounts/NRO-Cognitive-CustomVision-WestEurope-Prediction-S0";
var publication = _trainingClient.PublishIteration(projectId, iterationId, "Publication1", publicationResourceId);
// Force status refresh
targetIteration = _trainingClient.GetIteration(projectId, iterationId);
Console.WriteLine($"Iteration publish resource Id after publish: '{targetIteration.OriginalPublishResourceId}'");
}
}
}
See my Azure resource used:
I figured I'd upgrade my LuisRecognizer to use LuisRecognizerOptionsV3. However I can't seem to set prediction options the way I like - how do I set the timezone? The v3 prediction options lack this field.
In my bot I am currently doing:
var predictionOptions = new LuisPredictionOptions();
predictionOptions.TimezoneOffset = turnContext.Activity.LocalTimestamp.Value.Offset.TotalMinutes;
and I can't figure out the equivalent in v3's version of the data structure.
The timezoneOffset parameter was mostly provided as a way to determine what day it is for the user in case they say something like "today" or "tomorrow." It also helps when the user enters a relative time like "in three hours." When using the timezoneOffset parameter, the returned entity is in the provided timezone rather than universal time.
In LUIS v3, instead of providing an offset you provide a DateTime reference and LUIS uses that to process relative time. You can see that documented here. Note that the datetimeReference property is only available in POST requests and not GET requests because you provide it in the request body and not as a query parameter.
Also note that the datetimeReference property is not currently available in the Bot Builder SDK. You can write your own code to access the LUIS API directly with an HttpClient, but if you'd still like a prebuilt SDK to handle things then you can use this NuGet package: Microsoft.Azure.CognitiveServices.Language.LUIS.Runtime 3.0.0
Here's an example of how to use it:
var appId = new Guid("<LUIS APP ID>");
var client = new LUISRuntimeClient(new ApiKeyServiceClientCredentials("<SERVICE KEY>"));
client.Endpoint = "https://westus2.api.cognitive.microsoft.com";
var options = new PredictionRequestOptions(activity.LocalTimestamp.Value.DateTime);
var request = new PredictionRequest("Book a flight in three hours", options);
var response = await client.Prediction.GetSlotPredictionAsync(appId, "PRODUCTION", request);
Console.WriteLine(JsonConvert.SerializeObject(response.Prediction.Entities, Formatting.Indented));
I am using Google API to get information about an authenticated user. I can get the basic profile information, such as the ID and the full name. From the profile information, I can get the URL to the picture:
var plusMeUri = new Uri($"https://www.googleapis.com/plus/v1/people/me?key=<APP-ID>&access_token=<ACCESS-TOKEN>");
string userResponse = await HttpClient.GetStringAsync(plusMeUri);
JObject userObject = JObject.Parse(userResponse);
...
var imageObject = userObject.GetValue("image") as JObject;
var pictureUrl = imageObject.GetValue("url").Value<string>();
var pictureUri = new Uri(pictureUrl);
string uri = $"{pictureUri.Scheme}://{pictureUri.Host}{pictureUri.AbsolutePath}";
var pictureRequest = new HttpRequestMessage(HttpMethod.Get, uri);
pictureRequest.Headers.IfModifiedSince = <previous-timestamp>;
HttpResponseMessage pictureResponse = await HttpClient.SendAsync(pictureRequest);
if (pictureResponse.StatusCode == HttpStatusCode.NotModified)
// No need to handle anything else
return;
Question
I do not want to download the user's picture if it has not changed. This is why I am using the IfModifiedSince property. It does work with Facebook's API but it does not seem to work with Google's. How can I make it work?
From the information given, it seems like what you're trying to do is determine whether the image you're downloading/about to download is the same image as you've downloaded before. After looking at the Google+ API docs, it looks like the header you've been using isn't officially (at least not obviously) supported by their APIs.
But this is not the only way we can determine whether the image has changed or not (in fact, date last modified isn't necessarily the best way to do this anyway). Alternative methods include:
1) diffing the two images
2) checking the url (if we can assume different resources have different urls)
1 is likely the most accurate but also likely the least efficient, so I'll leave that to you to solve if you decide to go that route. I think the most promising is #2. I went ahead and played around with the API a little bit and it looks like the image.url field changes when you update your profile picture.
For example, here are my last two Google+ profile picture URLs:
https://lh4.googleusercontent.com/-oaUVPGFNkV8/AAAAAAAAAAI/AAAAAAAAAqs/KM7H8ZIFuxk/photo.jpg?sz=50
https://lh4.googleusercontent.com/-oaUVPGFNkV8/AAAAAAAAAAI/AAAAAAAAl24/yHU99opjgN4/photo.jpg?sz=50
As such, instead of waiting for the response from the server and checking its header to decide whether the image has been updated or not, you may be able to short-circuit the entire HTTP request by simply checking whether the last image you pulled down was from the same url or not. If it was from the same URL, it's likely you've already acquired that image otherwise you may not have it so should incur the cost of downloading anyway.
In this case, your code would read something like:
var imageObject = userObject.GetValue("image") as JObject;
var pictureUrl = imageObject.GetValue("url").Value<string>();
if(pictureUrl != <previous-picture-url>)
{
// insert get new picture logic here...
}
Anybody know where I can find a simple example C# code example? Apparently really tough to find.
I'm just starting out, got my Developer key.
Initial (really noob question/presumption) - -Can (should/must) my solution be a web service client? No new libraries I need to install in .Net right?
Basically, as a test, I want to be able to securely present a single note from a private notebook in html similar to what the Everfort export in html looks like on a outside WebSite.
Many Thanks in Advance!
You should start by downloading our API ZIP from http://www.evernote.com/about/developer/api/. You'll find C# client sample code in /sample/csharp. This sample code demonstrates using the Evernote API from a desktop application that authenticates using username and password.
I am not sure if you ever got this working, but I was playing around with Evernote, OpenAuth and C# this morning and managed to get it all working. I have put together a blog post / library explaining the experience and outlining how to do it with MVC here - http://www.shaunmccarthy.com/evernote-oauth-csharp/ - it uses the AsyncOAuth library: https://github.com/neuecc/AsyncOAuth
I wrote a wrapper around AsyncOAuth that you might find useful here: https://github.com/shaunmccarthy/AsyncOAuth.Evernote.Simple
One prickly thing to be aware of - the Evernote Endpoints (/oauth and /OAuth.action) are case sensitive
// Download the library from https://github.com/shaunmccarthy/AsyncOAuth.Evernote.Simple
// Configure the Authorizer with the URL of the Evernote service,
// your key, and your secret.
var EvernoteAuthorizer = new EvernoteAuthorizer(
"https://sandbox.evernote.com",
"slyrp-1234", // Not my real id / secret :)
"7acafe123456badb123");
// First of all, get a request token from Evernote - this causes a
// webrequest from your server to Evernote.
// The callBackUrl is the URL you want the user to return to once
// they validate the app
var requestToken = EvernoteAuthorizer.GetRequestToken(callBackUrl);
// Persist this token, as we are going to redirect the user to
// Evernote to Authorize this app
Session["RequestToken"] = requestToken;
// Generate the Evernote URL that we will redirect the user to in
// order to
var callForwardUrl = EvernoteAuthorizer.BuildAuthorizeUrl(requestToken);
// Redirect the user (e.g. MVC)
return Redirect(callForwardUrl);
// ... Once the user authroizes the app, they get redirected to callBackUrl
// where we parse the request parameter oauth_validator and finally get
// our credentials
// null = they didn't authorize us
var credentials = EvernoteAuthorizer.ParseAccessToken(
Request.QueryString["oauth_verifier"],
Session["RequestToken"] as RequestToken);
// Example of how to use the credential with Evernote SDK
var noteStoreUrl = EvernoteCredentials.NotebookUrl;
var noteStoreTransport = new THttpClient(new Uri(noteStoreUrl));
var noteStoreProtocol = new TBinaryProtocol(noteStoreTransport);
var noteStore = new NoteStore.Client(noteStoreProtocol);
List<Notebook> notebooks = client.listNotebooks(EvernoteCredentials.AuthToken);
http://weblogs.asp.net/psteele/archive/2010/08/06/edamlibrary-evernote-library-for-c.aspx might help. As the author states it just bundles some and fixes some. Haven't tried it myself but thought I'd mention for a possibly easier way to get started. Possibly.
This might help too...found it using the Way Back Machine since the original blog site was offline.
https://www.evernote.com/pub/bluecockatoo/Evernote_API#b=bb2451c9-b5ff-49bb-9686-2144d984c6ba&n=c30bc4eb-cca4-4a36-ad44-1e255eeb26dd
The original blog post: http://web.archive.org/web/20090203134615/http://macrolinz.com/macrolinz/index.php/2008/12/
Scroll down and find the post from December 26 - "Get it while it's hot..."