{"id":2389,"date":"2021-06-15T09:00:00","date_gmt":"2021-06-15T01:00:00","guid":{"rendered":"https:\/\/blog.ailia.ai\/%e6%9c%aa%e5%88%86%e9%a1%9e\/axgazeestimation-machine-learning-model-for-gaze-tracking\/"},"modified":"2025-05-14T15:30:17","modified_gmt":"2025-05-14T07:30:17","slug":"axgazeestimation-machine-learning-model-for-gaze-tracking","status":"publish","type":"post","link":"https:\/\/blog.ailia.ai\/en\/tips-en\/axgazeestimation-machine-learning-model-for-gaze-tracking\/","title":{"rendered":"AxGazeEstimation : A Machine Learning Model for Estimating Gaze"},"content":{"rendered":"\n<h3 class=\"wp-block-heading\"><strong>Overview<\/strong><\/h3>\n\n\n\n<p><em>AxGazeEstimation&nbsp;<\/em>is a machine learning model developed by&nbsp;<a href=\"https:\/\/axinc.jp\/en\/\" target=\"_blank\" rel=\"noreferrer noopener\">ax Inc<\/a>. to detect the direction of gaze of a person from an input image.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/cdn.pixabay.com\/photo\/2015\/02\/13\/14\/33\/vintage-635244_1280.jpg\" alt=\"\"\/><figcaption class=\"wp-element-caption\">\u51fa\u5178\uff1a<a href=\"https:\/\/pixabay.com\/ja\/photos\/%E3%83%93%E3%83%B3%E3%83%86%E3%83%BC%E3%82%B8-%E5%A5%B3%E6%80%A7-%E5%B8%BD%E5%AD%90-635244\/\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/pixabay.com\/ja\/photos\/%E3%83%93%E3%83%B3%E3%83%86%E3%83%BC%E3%82%B8-%E5%A5%B3%E6%80%A7-%E5%B8%BD%E5%AD%90-635244\/<\/a><\/figcaption><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\"><strong>Architecture<\/strong><\/h3>\n\n\n\n<p id=\"c851\"><em>AxGazeEstimation&nbsp;<\/em>uses&nbsp;<a href=\"https:\/\/medium.com\/axinc-ai\/blazeface-a-machine-learning-model-for-fast-detection-of-face-positions-and-key-points-5dcfb9429d72\"><em>BlazeFace<\/em><\/a><em>&nbsp;<\/em>to detect faces in an image and estimates the gaze using the detected face as input. Two methods of gaze estimation are available: direct estimation from the face image, and estimation from face image combined with face orientation.<\/p>\n\n\n\n<p id=\"3610\">The network backbone uses a reduced version of&nbsp;<em>ResNet50&nbsp;<\/em>(stage 3).<\/p>\n\n\n\n<p id=\"c851\">The training was performed using our in-house dataset made of 97,059 training images, and 11,775 validation images.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\"><strong>AxGazeEstimation Usage<\/strong><\/h3>\n\n\n\n<p>Use the following command to run the gaze estimation on the webcam video stream.<\/p>\n\n\n\n<p><div class=\"wp-block-code\"><code>$ python3 ax_gaze_estimation.py -v 0<\/code><\/div><\/p>\n\n\n\n<p>The following command can be used to estimate the face orientation in combination with the face detection.<\/p>\n\n\n\n<p><div class=\"wp-block-code\"><code>$ python3 ax_gaze_estimation.py -v 0 --include-head-pose<\/code><\/div><\/p>\n\n\n\n<p><a href=\"https:\/\/github.com\/axinc-ai\/ailia-models\/tree\/master\/face_recognition\/ax_gaze_estimation\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/github.com\/axinc-ai\/ailia-models\/tree\/master\/face_recognition\/ax_gaze_estimation<\/a><\/p>\n\n\n\n<p>Here is an example of&nbsp;<em>AxGazeEstimation&nbsp;<\/em>in action.<\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-rich is-provider-embed-handler wp-block-embed-embed-handler wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"wp-block-embed__wrapper\">\n<iframe loading=\"lazy\" title=\"ailia MODELS : AxGazeEstimation\" width=\"500\" height=\"281\" src=\"https:\/\/www.youtube.com\/embed\/qs79SakaPjI?feature=oembed\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe>\n<\/div><\/figure>\n\n\n\n<p id=\"8e43\"><a href=\"https:\/\/axinc.jp\/en\/\" rel=\"noreferrer noopener\" target=\"_blank\">ax Inc.<\/a>&nbsp;has developed&nbsp;<a href=\"https:\/\/ailia.jp\/en\/\" rel=\"noreferrer noopener\" target=\"_blank\">ailia SDK<\/a>, which enables cross-platform, GPU-based rapid inference.<\/p>\n\n\n\n<p id=\"7e07\">ax Inc. provides a wide range of services from consulting and model creation, to the development of AI-based applications and SDKs. Feel free to&nbsp;<a href=\"https:\/\/axinc.jp\/en\/\" rel=\"noreferrer noopener\" target=\"_blank\">contact us<\/a>&nbsp;for any inquiry.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Overview AxGazeEstimation&nbsp;is a machine learning model developed by&nbsp;ax Inc. to detect the direction of gaze of a person from an input image. Architecture AxGazeEstimation&nbsp;uses&nbsp;BlazeFace&nbsp;to detect faces in an image and estimates the gaze using the detected face as input. Two methods of gaze estimation are available: direct estimation from the face image, and estimation from face image combined with face orientation. The network backbone uses a reduced version of&nbsp;ResNet50&nbsp;(stage 3). The training was performed using our in-house dataset made of 97,059 training images, and 11,775 validation images. AxGazeEstimation Usage Use the following command to run the gaze estimation on the webcam video stream. $ python3 ax_gaze_estimation.py -v 0 The following [&hellip;]<\/p>\n","protected":false},"author":5,"featured_media":2110,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[255],"tags":[266],"class_list":["post-2389","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-tips-en","tag-ailiamodels-en"],"acf":[],"_links":{"self":[{"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/posts\/2389","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/users\/5"}],"replies":[{"embeddable":true,"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/comments?post=2389"}],"version-history":[{"count":3,"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/posts\/2389\/revisions"}],"predecessor-version":[{"id":2406,"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/posts\/2389\/revisions\/2406"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/media\/2110"}],"wp:attachment":[{"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/media?parent=2389"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/categories?post=2389"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blog.ailia.ai\/en\/wp-json\/wp\/v2\/tags?post=2389"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}