{"id":89110,"date":"2020-07-18T19:42:10","date_gmt":"2020-07-18T17:42:10","guid":{"rendered":"https:\/\/prohoster.info\/blog\/administrirovanie\/obshhij-obzor-arhitektury-servisa-dlya-oczenki-vneshnosti-na-osnove-nejronnyh-setej"},"modified":"2020-07-18T19:42:10","modified_gmt":"2020-07-18T17:42:10","slug":"obshhij-obzor-arhitektury-servisa-dlya-oczenki-vneshnosti-na-osnove-nejronnyh-setej","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/administrirovanie\/obshhij-obzor-arhitektury-servisa-dlya-oczenki-vneshnosti-na-osnove-nejronnyh-setej","title":{"rendered":"General overview of the service architecture for evaluating appearance based on neural networks","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

\"General<\/p>\n

Introduction<\/h2>\n

<\/p>\n

Hello!<\/p>\n

<\/p>\n

In this article, I will share my experience in building a microservices architecture for a project that uses neural networks.<\/p>\n

<\/p>\n

We will discuss the architecture requirements, take a look at various structural diagrams, break down each component of the completed architecture, and evaluate the technical metrics of the solution.<\/p>\n

<\/p>\n

Happy reading!<\/p>\n

<\/a><\/noindex><\/p>\n

A few words about the task and its solution<\/h2>\n

<\/p>\n

The main idea is to assess a person's attractiveness based on a photo using a ten-point scale.<\/p>\n

<\/p>\n

In this article, we will depart from discussing the neural networks used and the process of data preparation and training. However, in one of the upcoming publications, we will certainly return to a deeper analysis of the evaluation pipeline.<\/p>\n

<\/p>\n

For now, we will outline the evaluation pipeline at a high level, focusing on the interaction of microservices within the overall architecture of the project.\u00a0<\/p>\n

<\/p>\n

In working on the attractiveness evaluation pipeline, the task was decomposed into the following components:<\/p>\n

<\/p>\n

    \n
  1. Face detection in photos<\/li>\n
  2. Evaluation of each detected face<\/li>\n
  3. Rendering the result<\/li>\n<\/ol>\n

    <\/p>\n

    The first is handled by a pre-trained MTCNN<\/a><\/noindex>. For the second task, a convolutional neural network was trained on PyTorch, with ResNet34<\/a><\/noindex> being used as the backbone to balance 'quality \/ inference speed on CPU.'<\/p>\n

    \n\"General<\/p>\n

    Functional diagram of the evaluation pipeline<\/i><\/p>\n

    <\/p>\n

    Analysis of the architecture requirements for the project<\/h2>\n

    <\/p>\n

    In the lifecycle ML<\/a><\/noindex> of a project, the stages of working on the architecture and automating the deployment of the model are often among the most time-consuming and resource-intensive.<\/p>\n

    \n\"General<\/p>\n

    ML project lifecycle<\/i><\/p>\n

    <\/p>\n

    This project is no exception\u2014we decided to wrap the evaluation pipeline in an online service, which required delving into the architecture. The following basic requirements were outlined:<\/p>\n

    <\/p>\n

      \n
    1. A unified log storage\u2014 all services must write logs to a single location, which should be easy to analyze.<\/li>\n
    2. The ability to horizontally scale the evaluation service\u2014as the most likely bottleneck<\/li>\n
    3. Each image evaluation should allocate the same amount of CPU resources\u2014to avoid spikes in inference time distribution.<\/li>\n
    4. Fast (re)deployment of both specific services and the stack as a whole.<\/li>\n
    5. The ability to use shared objects in different services if necessary.<\/li>\n<\/ol>\n

      <\/p>\n

      Architecture<\/h2>\n

      <\/p>\n

      After analyzing the requirements, it became clear that microservice architecture fits almost perfectly.<\/p>\n

      <\/p>\n

      To avoid unnecessary headaches, Telegram API was chosen as the frontend.<\/p>\n

      <\/p>\n

      First, let's look at the structural diagram of the completed architecture, then we'll move on to describing each component and formalizing the process of successful image processing.<\/p>\n

      \n\"General<\/p>\n

      Structural diagram of the completed architecture<\/i><\/p>\n

      <\/p>\n

      Let's talk in more detail about each component of the diagram, clarifying their Single Responsibility in the image evaluation process.<\/p>\n

      <\/p>\n

      Microservice 'attrai-telegram-bot'<\/h3>\n

      <\/p>\n

      This microservice encapsulates all interactions with the Telegram API. There are two main scenarios to highlight \u2013 working with the user\u2019s image and handling the results of the evaluation pipeline. We'll discuss both scenarios in general.<\/p>\n

      <\/p>\n

      Upon receiving a user message with an image:<\/p>\n

      <\/p>\n

        \n
      1. Filtering is performed, consisting of the following checks:\n