Разработка клиентского приложения обнаружения объектов

Завершено

После обучения модели обнаружения объектов можно использовать пакет SDK пользовательского визуального распознавания Azure для разработки клиентского приложения, которое отправляет новые образы для анализа.

from msrest.authentication import ApiKeyCredentials
from azure.cognitiveservices.vision.customvision.prediction import CustomVisionPredictionClient


 # Authenticate a client for the prediction API
credentials = ApiKeyCredentials(in_headers={"Prediction-key": "<YOUR_PREDICTION_RESOURCE_KEY>"})
prediction_client = CustomVisionPredictionClient(endpoint="<YOUR_PREDICTION_RESOURCE_ENDPOINT>",
                                                 credentials=credentials)

# Get classification predictions for an image
image_data = open("<PATH_TO_IMAGE_FILE>", "rb").read()
results = prediction_client.detect_image("<YOUR_PROJECT_ID>",
                                           "<YOUR_PUBLISHED_MODEL_NAME>",
                                           image_data)

# Process predictions
for prediction in results.predictions:
    if prediction.probability > 0.5:
        left = prediction.bounding_box.left
        top = prediction.bounding_box.top 
        height = prediction.bounding_box.height
        width =  prediction.bounding_box.width
        print(f"{prediction.tag_name} ({prediction.probability})")
        print(f"  Left:{left}, Top:{top}, Height:{height}, Width:{width}")

using System;
using System.IO;
using Microsoft.Azure.CognitiveServices.Vision.CustomVision.Prediction;

// Authenticate a client for the prediction API
CustomVisionPredictionClient prediction_client = new CustomVisionPredictionClient(new ApiKeyServiceClientCredentials("<YOUR_PREDICTION_RESOURCE_KEY>"))
{
    Endpoint = "<YOUR_PREDICTION_RESOURCE_ENDPOINT>"
};

// Get classification predictions for an image
MemoryStream image_data = new MemoryStream(File.ReadAllBytes("<PATH_TO_IMAGE_FILE>"));
var result = prediction_client.DetectImage("<YOUR_PROJECT_ID>",
                                             "<YOUR_PUBLISHED_MODEL_NAME>",
                                             image_data);

// Process predictions
foreach (var prediction in result.Predictions)
{
    if (prediction.Probability > 0.5)
    {
        var left = prediction.BoundingBox.Left;
        var top = prediction.BoundingBox.Top;
        var height = prediction.BoundingBox.Height;
        var width =  prediction.BoundingBox.Width;
        Console.WriteLine($"{prediction.TagName} ({prediction.Probability})");
        Console.WriteLine($"  Left:{left}, Top:{top}, Height:{height}, Width:{width}");
    }
}