Blog

Verifying faces

When a face is detected by the Face service, a unique ID is assigned to it and retained in the service resource for 24 hours. The ID is a GUID, with no indication of the individual’s identity other than their facial features.

While the detected face ID is cached, subsequent images can be used to compare the new faces to the cached identity and determine if they’re similar (in other words, they share similar facial features) or to verify that the same person appears in two images.

This ability to compare faces anonymously can be useful in systems where it’s important to confirm that the same person is present on two occasions, without the need to know the actual identity of the person. For example, by taking images of people as they enter and leave a secured space to verify that everyone who entered leaves.

Identifying faces

For scenarios where you need to positively identify individuals, you can train a facial recognition model using face images.

To train a facial recognition model with the Face service:

1. Create a Person Group that defines the set of individuals you want to identify (for example, employees).
2. Add a Person to the Person Group for each individual you want to identify.
3. Add detected faces from multiple images to each person, preferably in various poses. The IDs of these faces will no longer expire after 24 hours (so they’re now referred to as persisted faces).
4. Train the model.

staff augmentation service

To use the Azure AI Vision Face API, you must provision a resource for the service in an Azure subscription. You can provision Face as a single-service resource, or you can use the Face API in a multi-service Azure AI Services resource; which can be provisioned as a standalone resource or as part of an Azure AI Foundry hub.

To use your resource from a client application you must connect to its endpoint using either key-based authentication or Microsoft Entra AI authentication. When using the REST interface you can provide the authentication key or token in the request header. When using a language-specific SDK (for example, the Python azure-ai-vision-face package or the Microsoft .NET Azure.AI.Vision.Face package), you use a FaceClient object to connect to the service.

software development

The Face service provides functionality that you can use for:

1. Face detection – for each detected face, the results include an ID that identifies the face and the bounding box coordinates indicating its location in the image.
2. Face attribute analysis – you can return a wide range of facial attributes, including:
   • Head pose (pitch, roll, and yaw orientation in 3D space)
   • Glasses (No glasses, Reading glasses, Sunglasses, or Swimming Goggles)
   • Mask (the presence of a face mask)
   • Blur (low, medium, or high)
   • Exposure (under exposure, good exposure, or over exposure)
   • Noise (visual noise in the image)
   • Occlusion (objects obscuring the face)
   • Accessories (glasses, headwear, mask)
   • QualityForRecognition (low, medium, or high)
3. Facial landmark location – coordinates for key landmarks in relation to facial features (for example, eye corners, pupils, tip of nose, and so on)
4. Face comparison – you can compare faces across multiple images for similarity (to find individuals with similar facial features) and verification (to determine that a face in one image is the same person as a face in another image)
5. Facial recognition – you can train a model with a collection of faces belonging to specific individuals, and use the model to identify those people in new images.
6. Facial liveness – liveness can be used to determine if the input video is a real stream or a fake to prevent bad-intentioned individuals from spoofing a facial recognition system.

school management

Automated evaluations in Azure AI Foundry portal enable you to assess the quality and content safety performance of models, datasets, or prompt flows.

Evaluation data

To evaluate a model, you need a dataset of prompts and responses (and optionally, expected responses as “ground truth”). You can compile this dataset manually or use the output from an existing application; but a useful way to get started is to use an AI model to generate a set of prompts and responses related to a specific subject. You can then edit the generated prompts and responses to reflect your desired output, and use them as ground truth to evaluate the responses from another model.

Evaluation metrics

Automated evaluation enables you to choose which evaluators you want to assess your model’s responses, and which metrics those evaluators should calculate. There are evaluators that help you measure:

• AI Quality: The quality of your model’s responses is measured by using AI models to evaluate them for metrics like coherence and relevance and by using standard NLP metrics like F1 score, BLEU, METEOR, and ROUGE based on ground truth (in the form of expected response text)
• Risk and safety: evaluators that assess the responses for content safety issues, including violence, hate, sexual content, and content related to self-harm.

quote

During the early phases of the development of your generative AI app, you want to experiment and iterate quickly. To easily assess whether your selected language model and app, created with prompt flow, meet your requirements, you can manually evaluate models and flows in the Azure AI Foundry portal.

Even when your model and app are already in production, manual evaluations are a crucial part of assessing performance. As manual evaluations are done by humans, they can provide insights that automated metrics might miss.

Let’s explore how you can manually evaluate your selected models and app in the Azure AI Foundry portal.

Prepare your test prompts

To begin the manual evaluation process, it’s essential to prepare a diverse set of test prompts that reflect the range of queries and tasks your app is expected to handle. These prompts should cover various scenarios, including common user questions, edge cases, and potential failure points. By doing so, you can comprehensively assess the app’s performance and identify areas for improvement.

mobile app development

Evaluating your model’s performance at different phases is crucial to ensure its effectiveness and reliability. Before exploring the various options you have to evaluate your model, let’s explore the aspects of your application you can evaluate.

When you develop a generative AI app, you use a language model in your chat application to generate a response. To help you decide which model you want to integrate into your application, you can evaluate the performance of an individual language model:

An input (1) is provided to a language model (2), and a response is generated as output (3). The model is then evaluated by analyzing the input, the output, and optionally comparing it to predefined expected output.

managed it services

hile you can perform all video analysis tasks in the Azure Video Indexer portal, you may want to incorporate the service into custom applications. There are two ways you can accomplish this.

Azure Video Indexer widgets

The widgets used in the Azure Video Indexer portal to play, analyze, and edit videos can be embedded in your own custom HTML interfaces. You can use this technique to share insights from specific videos with others without giving them full access to your account in the Azure Video Indexer portal.

Azure Video Indexer API

Azure Video Indexer provides a REST API that you can use to obtain information about your account, including an access token.

HTTPCopy

https://api.videoindexer.ai/Auth/<location>/Accounts/<accountId>/AccessToken

You can then use your token to consume the REST API and automate video indexing tasks, creating projects, retrieving insights, and creating or deleting custom models.

For example, a GET call to https://api.videoindexer.ai/<location>/Accounts/<accountId>/Customization/CustomLogos/Logos/<logoId>?<accessToken> REST endpoint returns the specified logo. In another example, you can send a GET request to https://api.videoindexer.ai/<location>/Accounts/<accountId>/Videos?<accessToken>, which returns details of videos in your account, similar to the following JSON example:

JSONCopy

{
    "accountId": "SampleAccountId",
    "id": "30e66ec1b1",
    "partition": null,
    "externalId": null,
    "metadata": null,
    "name": "test3",
    "description": null,
    "created": "2018-04-25T16=50=00.967+00=00",
    "lastModified": "2018-04-25T16=58=13.409+00=00",
    "lastIndexed": "2018-04-25T16=50=12.991+00=00",
    "privacyMode": "Private",
    "userName": "SampleUserName",
    "isOwned": true,
    "isBase": true,
    "state": "Processing",
    "processingProgress": "",
    "durationInSeconds": 13,
    "thumbnailVideoId": "30e66ec1b1",
    "thumbnailId": "55848b7b-8be7-4285-893e-cdc366e09133",
    "social": {
        "likedByUser": false,
        "likes": 0,
        "views": 0
    },
    "searchMatches": [],
    "indexingPreset": "Default",
    "streamingPreset": "Default",
    "sourceLanguage": "en-US"
}

loyalty

Azure Video Indexer includes predefined models that can recognize well-known celebrities, do OCR, and transcribe spoken phrases into text. You can extend the recognition capabilities of Video Analyzer by creating custom models for:

• People. Add images of the faces of people you want to recognize in videos, and train a model. Video Indexer will then recognize these people in all of your videos.

• Language. If your organization uses specific terminology that may not be in common usage, you can train a custom model to detect and transcribe it.
• Brands. You can train a model to recognize specific names as brands, for example to identify products, projects, or companies that are relevant to your business.

it support

The Azure Video Indexer service is designed to help you extract information from videos. It provides functionality that you can use for:

• Facial recognition – detecting the presence of individual people in the image. This requires Limited Access approval.
• Optical character recognition – reading text in the video.
• Speech transcription – creating a text transcript of spoken dialog in the video.
• Topics – identification of key topics discussed in the video.
• Sentiment – analysis of how positive or negative segments within the video are.
• Labels – label tags that identify key objects or themes throughout the video.
• Content moderation – detection of adult or violent themes in the video.
• Scene segmentation – a breakdown of the video into its constituent scenes.

it consulting

A cmdlet (pronounced “command-let”) is a compiled command. A cmdlet can be developed in .NET or .NET Core and invoked as a command within PowerShell. Thousands of cmdlets are available in your PowerShell installation. The challenge lies in discovering what the cmdlets are and what they can do for you.

Cmdlets are named according to a verb-noun naming standard. This pattern can help you to understand what they do and how to search for them. It also helps cmdlet developers create consistent names. You can see the list of approved verbs by using the Get-Verb cmdlet. Verbs are organized according to activity type and function.

infrastructure services