Cloudinary Blog

Powerful image manipulation and categorization with facial attribute detection

Advanced Facial Attributes Detection for Image Manipulation

Update - December 2015: The add-on described in this post is no longer available since ReKognition terminated their services. However, all features described here are still available via a different and even better add-on: Advanced Facial Attributes Detection

Face Detection is a great feature that enables the automatic modification of images according to the detected faces within an image, making it simple to intelligently crop, position, resize and transform your images appropriately.

Facial Attribute Detection takes the process even further and extracts meaningful advanced data about the face(s) in the image, including the exact location of facial features. This allows you even greater control over your image categorization, and to automatically use these details to smartly crop, position, rotate and overlay images based on the detected facial features.

Facial Attribute Detection lets you know more than just the location of a person's facial features. How are they posed in 3 dimensions? Is the person wearing glasses? Do they have their eyes closed? Mouth open? Have a mustache or beard? What is the person's race, age and gender? What emotion are they displaying? Are they smiling? How beautiful are they? Retrieving this information makes it a simple matter to automatically categorize and tag your collection of images.

All of this is made possible with the ReKognition add-on, which has been directly integrated within Cloudinary’s infrastructure, further extending Cloudinary’s built-in face detection to a robust Facial Attribute Detection feature. By simply setting the detection parameter to rekognition_face when calling Cloudinary's upload API, ReKognition is utilized to automatically extract detailed face attributes from the uploaded image.

Ruby:
Cloudinary::Uploader.upload("woman.jpg", 
              :detection => "rekognition_face")
PHP:
\Cloudinary\Uploader::upload("woman.jpg", 
              array(
               "detection" => "rekognition_face"));
Python:
cloudinary.uploader.upload("woman.jpg", 
              detection = "rekognition_face")
Node.js:
cloudinary.uploader.upload("woman.jpg", 
              function(result) {console.log(result); }, { detection: "rekognition_face" });
Java:
cloudinary.uploader().upload("woman.jpg", 
              Cloudinary.asMap("detection", "rekognition_face"));

Original woman image

The example JSON snippet displayed from the example image above contains the result returned from the face ReKognition request, which includes very detailed information regarding the face that was automatically detected in the image.

{"rekognition_face": 
 "status": "complete",
  "data": [
    {
      "boundingbox": {
        "tl": {"x": 231.45, "y": 102.52},
        "size": {"width": 240.77, "height": 240.77 }},
      "confidence": 1,
      "eye_left": {"x":309.6, "y": 190.1},
      "eye_right": {"x": 407.9, "y": 213.6},
      "nose": {"x": 199.1, "y": 204.0},
      
      
      
      "smile": 0.96,
      "glasses": 0.01,
      "sunglasses": 0.04,
      "beard": 0,
      "mustache": 0,
      "eye_closed": 0,
      "mouth_open_wide": 0.73,
      "beauty": 0.63531,
      "sex": 1
    }
  ]
}

You can also use Cloudinary's Admin API to apply ReKognition face detection to already uploaded images (based on their public IDs), and the face attributes that were previously extracted are also available using the Admin API's show resource details method.

Face detection based cropping

Based on the position of facial attributes detected by the ReKognition add-on, Cloudinary can crop your images to focus on the detected facial features, while providing a large set of image transformation and cropping options when using a Cloudinary delivery URL or calling Cloudinary's image API.

To focus an automatic crop on the detected faces, simply set the crop parameter to thumb, fill or crop and the gravity parameter to rek_faces (set gravity to rek_face for focusing on the single largest detected face in the image). The resulting images are dynamically generated on-the-fly and the result is delivered via a fast CDN.

Original photo

The following code sample generates a 150x150 thumbnail of the nice_coupleimage shown above, using multiple face detection based cropping.

Ruby:
cl_image_tag("nice_couple.jpg", :gravity=>"rek_faces", :width=>150, :height=>150, :crop=>"thumb")
PHP:
cl_image_tag("nice_couple.jpg", array("gravity"=>"rek_faces", "width"=>150, "height"=>150, "crop"=>"thumb"))
Python:
CloudinaryImage("nice_couple.jpg").image(gravity="rek_faces", width=150, height=150, crop="thumb")
Node.js:
cloudinary.image("nice_couple.jpg", {gravity: "rek_faces", width: 150, height: 150, crop: "thumb"})
Java:
cloudinary.url().transformation(new Transformation().gravity("rek_faces").width(150).height(150).crop("thumb")).imageTag("nice_couple.jpg");
JS:
cloudinary.imageTag('nice_couple.jpg', {gravity: "rek_faces", width: 150, height: 150, crop: "thumb"}).toHtml();
jQuery:
$.cloudinary.image("nice_couple.jpg", {gravity: "rek_faces", width: 150, height: 150, crop: "thumb"})
React:
<Image publicId="nice_couple.jpg" >
  <Transformation gravity="rek_faces" width="150" height="150" crop="thumb" />
</Image>
Angular:
<cl-image public-id="nice_couple.jpg" >
  <cl-transformation gravity="rek_faces" width="150" height="150" crop="thumb">
  </cl-transformation>
</cl-image>
.Net:
cloudinary.Api.UrlImgUp.Transform(new Transformation().Gravity("rek_faces").Width(150).Height(150).Crop("thumb")).BuildImageTag("nice_couple.jpg")
Android:
MediaManager.get().url().transformation(new Transformation().gravity("rek_faces").width(150).height(150).crop("thumb")).generate("nice_couple.jpg");
iOS:
imageView.cldSetImage(cloudinary.createUrl().setTransformation(CLDTransformation().setGravity("rek_faces").setWidth(150).setHeight(150).setCrop("thumb")).generate("nice_couple.jpg")!, cloudinary: cloudinary)
150x150 thumbnail of nice_couple.jpg

Eyes detection based cropping

Cloudinary can also dynamically crop your images based on the position of detected eyes. Simply set the gravity parameter to rek_eyes (g_rek_eyes for URLs) to center the image on the detected eyes. The example below delivers a 200x60 thumbnail centered on the eyes:

Ruby:
cl_image_tag("woman.jpg", :gravity=>"rek_eyes", :width=>200, :height=>60, :crop=>"thumb")
PHP:
cl_image_tag("woman.jpg", array("gravity"=>"rek_eyes", "width"=>200, "height"=>60, "crop"=>"thumb"))
Python:
CloudinaryImage("woman.jpg").image(gravity="rek_eyes", width=200, height=60, crop="thumb")
Node.js:
cloudinary.image("woman.jpg", {gravity: "rek_eyes", width: 200, height: 60, crop: "thumb"})
Java:
cloudinary.url().transformation(new Transformation().gravity("rek_eyes").width(200).height(60).crop("thumb")).imageTag("woman.jpg");
JS:
cloudinary.imageTag('woman.jpg', {gravity: "rek_eyes", width: 200, height: 60, crop: "thumb"}).toHtml();
jQuery:
$.cloudinary.image("woman.jpg", {gravity: "rek_eyes", width: 200, height: 60, crop: "thumb"})
React:
<Image publicId="woman.jpg" >
  <Transformation gravity="rek_eyes" width="200" height="60" crop="thumb" />
</Image>
Angular:
<cl-image public-id="woman.jpg" >
  <cl-transformation gravity="rek_eyes" width="200" height="60" crop="thumb">
  </cl-transformation>
</cl-image>
.Net:
cloudinary.Api.UrlImgUp.Transform(new Transformation().Gravity("rek_eyes").Width(200).Height(60).Crop("thumb")).BuildImageTag("woman.jpg")
Android:
MediaManager.get().url().transformation(new Transformation().gravity("rek_eyes").width(200).height(60).crop("thumb")).generate("woman.jpg");
iOS:
imageView.cldSetImage(cloudinary.createUrl().setTransformation(CLDTransformation().setGravity("rek_eyes").setWidth(200).setHeight(60).setCrop("thumb")).generate("woman.jpg")!, cloudinary: cloudinary)
200x60 thumbnail centered on eyes

Facial overlays

Thanks to the detailed information on the position of facial attributes detected by ReKognition, Cloudinary can add overlays while taking into account the pose of the face, and automatically scale and rotate the overlay accordingly.

Ruby:
cl_image_tag("HarlequinMask.jpg")
PHP:
cl_image_tag("HarlequinMask.jpg")
Python:
CloudinaryImage("HarlequinMask.jpg").image()
Node.js:
cloudinary.image("HarlequinMask.jpg")
Java:
cloudinary.url().imageTag("HarlequinMask.jpg");
JS:
cloudinary.imageTag('HarlequinMask.jpg').toHtml();
jQuery:
$.cloudinary.image("HarlequinMask.jpg")
React:
<Image publicId="HarlequinMask.jpg" >

</Image>
Angular:
<cl-image public-id="HarlequinMask.jpg" >

</cl-image>
.Net:
cloudinary.Api.UrlImgUp.BuildImageTag("HarlequinMask.jpg")
Android:
MediaManager.get().url().generate("HarlequinMask.jpg");
iOS:
imageView.cldSetImage(cloudinary.createUrl().generate("HarlequinMask.jpg")!, cloudinary: cloudinary)
Harlequin mask

For example, in order to automatically overlay the above image of a harlequin mask scaled to 160% relative to the detected eyes in the main image:

Ruby:
cl_image_tag("woman.jpg", :flags=>"region_relative", :gravity=>"rek_eyes", :overlay=>"HarlequinMask", :width=>1.6, :crop=>"scale")
PHP:
cl_image_tag("woman.jpg", array("flags"=>"region_relative", "gravity"=>"rek_eyes", "overlay"=>"HarlequinMask", "width"=>1.6, "crop"=>"scale"))
Python:
CloudinaryImage("woman.jpg").image(flags="region_relative", gravity="rek_eyes", overlay="HarlequinMask", width=1.6, crop="scale")
Node.js:
cloudinary.image("woman.jpg", {flags: "region_relative", gravity: "rek_eyes", overlay: "HarlequinMask", width: "1.6", crop: "scale"})
Java:
cloudinary.url().transformation(new Transformation().flags("region_relative").gravity("rek_eyes").overlay(new Layer().publicId("HarlequinMask")).width(1.6).crop("scale")).imageTag("woman.jpg");
JS:
cloudinary.imageTag('woman.jpg', {flags: "region_relative", gravity: "rek_eyes", overlay: new cloudinary.Layer().publicId("HarlequinMask"), width: "1.6", crop: "scale"}).toHtml();
jQuery:
$.cloudinary.image("woman.jpg", {flags: "region_relative", gravity: "rek_eyes", overlay: new cloudinary.Layer().publicId("HarlequinMask"), width: "1.6", crop: "scale"})
React:
<Image publicId="woman.jpg" >
  <Transformation flags="region_relative" gravity="rek_eyes" overlay="HarlequinMask" width="1.6" crop="scale" />
</Image>
Angular:
<cl-image public-id="woman.jpg" >
  <cl-transformation flags="region_relative" gravity="rek_eyes" overlay="HarlequinMask" width="1.6" crop="scale">
  </cl-transformation>
</cl-image>
.Net:
cloudinary.Api.UrlImgUp.Transform(new Transformation().Flags("region_relative").Gravity("rek_eyes").Overlay(new Layer().PublicId("HarlequinMask")).Width(1.6).Crop("scale")).BuildImageTag("woman.jpg")
Android:
MediaManager.get().url().transformation(new Transformation().flags("region_relative").gravity("rek_eyes").overlay(new Layer().publicId("HarlequinMask")).width(1.6).crop("scale")).generate("woman.jpg");
iOS:
imageView.cldSetImage(cloudinary.createUrl().setTransformation(CLDTransformation().setFlags("region_relative").setGravity("rek_eyes").setOverlay("HarlequinMask").setWidth(1.6).setCrop("scale")).generate("woman.jpg")!, cloudinary: cloudinary)
Harlequin masked face

Heres another example, this time with glasses.

Ruby:
cl_image_tag("glasses.jpg")
PHP:
cl_image_tag("glasses.jpg")
Python:
CloudinaryImage("glasses.jpg").image()
Node.js:
cloudinary.image("glasses.jpg")
Java:
cloudinary.url().imageTag("glasses.jpg");
JS:
cloudinary.imageTag('glasses.jpg').toHtml();
jQuery:
$.cloudinary.image("glasses.jpg")
React:
<Image publicId="glasses.jpg" >

</Image>
Angular:
<cl-image public-id="glasses.jpg" >

</cl-image>
.Net:
cloudinary.Api.UrlImgUp.BuildImageTag("glasses.jpg")
Android:
MediaManager.get().url().generate("glasses.jpg");
iOS:
imageView.cldSetImage(cloudinary.createUrl().generate("glasses.jpg")!, cloudinary: cloudinary)
Glasses

Overlaying the above image scaled to 150% relative to the detected eyes in the main image, which is then presented as a 200 pixel wide round thumbnail centered on the face:

Ruby:
cl_image_tag("woman.jpg", :transformation=>[
  {:flags=>"region_relative", :gravity=>"rek_eyes", :overlay=>"glasses", :width=>1.5, :crop=>"scale"},
  {:width=>200, :gravity=>"face", :radius=>"max", :crop=>"thumb"}
  ])
PHP:
cl_image_tag("woman.jpg", array("transformation"=>array(
  array("flags"=>"region_relative", "gravity"=>"rek_eyes", "overlay"=>"glasses", "width"=>1.5, "crop"=>"scale"),
  array("width"=>200, "gravity"=>"face", "radius"=>"max", "crop"=>"thumb")
  )))
Python:
CloudinaryImage("woman.jpg").image(transformation=[
  {'flags': "region_relative", 'gravity': "rek_eyes", 'overlay': "glasses", 'width': 1.5, 'crop': "scale"},
  {'width': 200, 'gravity': "face", 'radius': "max", 'crop': "thumb"}
  ])
Node.js:
cloudinary.image("woman.jpg", {transformation: [
  {flags: "region_relative", gravity: "rek_eyes", overlay: "glasses", width: "1.5", crop: "scale"},
  {width: 200, gravity: "face", radius: "max", crop: "thumb"}
  ]})
Java:
cloudinary.url().transformation(new Transformation()
  .flags("region_relative").gravity("rek_eyes").overlay(new Layer().publicId("glasses")).width(1.5).crop("scale").chain()
  .width(200).gravity("face").radius("max").crop("thumb")).imageTag("woman.jpg");
JS:
cloudinary.imageTag('woman.jpg', {transformation: [
  {flags: "region_relative", gravity: "rek_eyes", overlay: new cloudinary.Layer().publicId("glasses"), width: "1.5", crop: "scale"},
  {width: 200, gravity: "face", radius: "max", crop: "thumb"}
  ]}).toHtml();
jQuery:
$.cloudinary.image("woman.jpg", {transformation: [
  {flags: "region_relative", gravity: "rek_eyes", overlay: new cloudinary.Layer().publicId("glasses"), width: "1.5", crop: "scale"},
  {width: 200, gravity: "face", radius: "max", crop: "thumb"}
  ]})
React:
<Image publicId="woman.jpg" >
  <Transformation flags="region_relative" gravity="rek_eyes" overlay="glasses" width="1.5" crop="scale" />
  <Transformation width="200" gravity="face" radius="max" crop="thumb" />
</Image>
Angular:
<cl-image public-id="woman.jpg" >
  <cl-transformation flags="region_relative" gravity="rek_eyes" overlay="glasses" width="1.5" crop="scale">
  </cl-transformation>
  <cl-transformation width="200" gravity="face" radius="max" crop="thumb">
  </cl-transformation>
</cl-image>
.Net:
cloudinary.Api.UrlImgUp.Transform(new Transformation()
  .Flags("region_relative").Gravity("rek_eyes").Overlay(new Layer().PublicId("glasses")).Width(1.5).Crop("scale").Chain()
  .Width(200).Gravity("face").Radius("max").Crop("thumb")).BuildImageTag("woman.jpg")
Android:
MediaManager.get().url().transformation(new Transformation()
  .flags("region_relative").gravity("rek_eyes").overlay(new Layer().publicId("glasses")).width(1.5).crop("scale").chain()
  .width(200).gravity("face").radius("max").crop("thumb")).generate("woman.jpg");
iOS:
imageView.cldSetImage(cloudinary.createUrl().setTransformation(CLDTransformation()
  .setFlags("region_relative").setGravity("rek_eyes").setOverlay("glasses").setWidth(1.5).setCrop("scale").chain()
  .setWidth(200).setGravity("face").setRadius("max").setCrop("thumb")).generate("woman.jpg")!, cloudinary: cloudinary)
Glasses overlayed on eyes

Summary

The ReKognition add-on is utilized to automatically extract detailed face attributes from your images, and enables advanced image manipulation and categorization based on the detected facial data, with relative scaling and rotation of overlays achieved automatically.

ReKognition detect face attributes add-on

The ReKognition add-on is available to all our free and paid plans. If you don't have a Cloudinary account, you are welcome to sign up to our free account and try it out.

Update - December 2015: The add-on described in this post is no longer available since ReKognition terminated their services. However, all features described here are still available via a different and even better add-on: Advanced Facial Attributes Detection

Recent Blog Posts

Hipcamp Optimizes Images and Improves Page Load Times With Cloudinary

When creating a website that allows campers to discover great destinations, Hipcamp put a strong emphasis on featuring high-quality images that showcased the list of beautiful locations, regardless of whether users accessed the site on a desktop, tablet, or phone. Since 2015, Hipcamp has relied on Cloudinary’s image management solution to automate cropping and image optimization, enabling instant public delivery of photos, automatic tagging based on content recognition, and faster loading of webpages. In addition, Hipcamp was able to maintain the high standards it holds for the look and feel of its website.

Read more
New Image File Format: FUIF: Why Do We Need a New Image Format

In my last post, I introduced FUIF, a new, free, and universal image format I’ve created. In this post and other follow-up pieces, I will explain the why, what, and how of FUIF.

Even though JPEG is still the most widely-used image file format on the web, it has limitations, especially the subset of the format that has been implemented in browsers and that has, therefore, become the de facto standard. Because JPEG has a relatively verbose header, it cannot be used (at least not as is) for low-quality image placeholders (LQIP), for which you need a budget of a few hundred bytes. JPEG cannot encode alpha channels (transparency); it is restricted to 8 bits per channel; and its entropy coding is no longer state of the art. Also, JPEG is not fully “responsive by design.” There is no easy way to find a file’s truncation offsets and it is limited to a 1:8 downscale (the DC coefficients). If you want to use the same file for an 8K UHD display (7,680 pixels wide) and for a smart watch (320 pixels wide), 1:8 is not enough. And finally, JPEG does not work well with nonphotographic images and cannot do fully lossless compression.

Read more
 New Image File Format: FUIF:Lossy, Lossless, and Free

I've been working to create a new image format, which I'm calling FUIF, or Free Universal Image Format. That’s a rather pretentious name, I know. But I couldn’t call it the Free Lossy Image Format (FLIF) because that acronym is not available any more (see below) and FUIF can do lossless, too, so it wouldn’t be accurate either.

Read more
Optimizing Video Streaming and Delivery: Q&A with Doug Sillars

Doug Sillars, a digital nomad and a freelance mobile-performance expert, answers questions about video streaming and delivery, website optimization, and more.

Doug Sillars, a freelance mobile-performance expert and developer advocate, is a Google Developer Expert and the author of O’Reilly’s High Performance Android Apps. Given his extensive travels across the globe—from the UK to Siberia—with his wife, kids, and 11-year-old dog, Max, he has been referred to as a “digital nomad.” So far in 2018, Doug has spoken at more than 75 meetups and conferences!

Read more