{"id":21863,"date":"2020-08-23T10:35:00","date_gmt":"2020-08-23T10:35:00","guid":{"rendered":"http:\/\/how_to_automatically_and_professionally_remove_photo_backgrounds"},"modified":"2025-05-03T15:31:45","modified_gmt":"2025-05-03T22:31:45","slug":"how_to_automatically_and_professionally_remove_photo_backgrounds","status":"publish","type":"post","link":"https:\/\/cloudinary.com\/blog\/how_to_automatically_and_professionally_remove_photo_backgrounds","title":{"rendered":"How to Automatically Remove Photo Backgrounds in Seconds With AI"},"content":{"rendered":"

These days, e-commerce websites are trending clean and sleek, almost always spotlighting high-quality product photos with a flat-white, solid-color, or transparent background. Many practical and well-validated reasons support this choice, making it clear that this is no passing fad. This graphic illustrates a few of them:<\/p>\n

\"Infographic:<\/p>\n

How do website or app developers deliver hundreds or thousands of solid- or transparent-background images, including user-generated content? Getting someone to manually edit<\/a> them with an editing tool definitely doesn\u2019t scale.<\/p>\n

Cloudinary<\/a> has the solution in its AI Background-Removal Add-On<\/strong>. The add-on\u2019s deep-learning and artificial-intelligence algorithms identify the primary subject in each photo\u2019s foreground and then accurately remove its background in mere seconds.<\/p>\n

Moments after you upload your photos, \u2026<\/p>\n

\n\n \n \n \n
\n These . . .<\/strong>\n <\/td>\n \n Become These!<\/strong>\n <\/td>\n <\/tr>\n
\n\"Dog<\/a>\n <\/td>\n \"Dog<\/a>\n <\/td>\n <\/tr>\n
\"Baseball<\/a>\n <\/td>\n <\/td>\n \"Baseball<\/a>\n <\/td>\n <\/tr>\n<\/table>\n<\/center>\n

Transparent Backgrounds Are Just a Parameter Away<\/h2>\n

Cloudinary is a robust media-management platform that leverages automation and deep-learning algorithms to simplify and accelerate workflows for website and app developers, from bulk uploads of raw photos to on-the-fly transformations for creating multiple media variants, optimizations, and responsive delivery to all devices.<\/p>\n

To apply the AI Background Removal Add-On to images during or after uploading them to your Cloudinary account, add the background_removal<\/code> parameter set to cloudinary_ai<\/code> in your upload<\/code> or update<\/code> command. When triggered in the method call, the add-on analyzes the image and then replaces the originally uploaded image with a transparent PNG containing only the foreground image. Cloudinary backs up the original for later retrieval, if required.For example, to have Cloudinary replace the the baseball JPG<\/a> image above with a transparent PNG, you could run the following upload command:<\/p>\n\n Loading code examples<\/span>\n <\/cld-code-widget>\n

Even though background removal<\/a> takes only a few seconds, if you plan to deliver the transparent image on your site or app immediately after upload, it\u2019s a good idea to include the notification_url<\/code> parameter in your upload<\/code> or upload<\/code> command, which activates a webhook that receives the removal status. You can then check if removal is still in progress and, if so, you might want to display a placeholder image or an in-progress animation until the process is complete.<\/p>\n

We Only Make It Look Simple<\/h2>\n

Segmenting objects from images is a fundamental and well-studied challenge in the computer industry. Nowadays, neural networks are the \u201csecret sauce\u201d behind many segmentation<\/a> tasks that enable state-of-the-art functionality for various applications, such as autonomous driving, medical analysis, and industrial planning.<\/p>\n

As shown above, erasing image backgrounds with Cloudinary\u2019s add-on is simple for the developers who use it.. Behind the scenes, however, the Cloudinary engine must perform these three complex tasks::<\/p>\n

    \n
  1. \n

    Identify the salient objects in the image to determine which objects to classify as foreground<\/strong> according to the context and scene composition<\/a>.<\/p>\n<\/li>\n

  2. \n

    Segment the salient objects. Most neutral networks today produce only coarse maps through standard segmentation techniques. For images slated for production, the segmentation maps<\/strong> of foreground versus background pixels must be near-perfect to achieve seamless background removal, e.g., proper segmentation of fur and hair requires meticulous processing.<\/p>\n

    Currently, no large public<\/strong> dataset exists for accurate background removal, rendering it all the more difficult to acquire the quantity of media assets necessary for training neutral networks.<\/p>\n<\/li>\n

  3. \n

    Separate the foreground as an alpha layer.<\/p>\n

    While addressing those challenges, we at Cloudinary also focused on finding ways to optimize<\/strong> our algorithm to deliver the high performance our customers expect; that means processing images of all sizes at scale.<\/p>\n<\/li>\n<\/ol>\n

    Optimal Results Are Paramount<\/h2>\n

    When you prepare the delivery URLs for the resulting transparent PNGs, you might want to do one or both of the following:<\/p>\n