An Image Editing story
Images are the most powerful tool in online retail (a.k.a e-tail). Occupying over 75% of any website page, images are the first impression that registers 50,000 times faster than any written content. Naturally, images are also resource intensive & costly to produce.
For any large e-tailer, this is a nightmare. With multiple studio’s, photographers, stylists, expensive gear & models, optimizing photography is like finding the holy grail.
However, post-production of images is a different story. It is mostly a repetitive job, especially clipping of images, changing the background and basic retouching of model and/or products. An industry with high attrition rates, it is almost impossible to maintain quality & speed. Often, a photo-studio manage to shoot 100’s of products in a day while their post-production takes many days. And then there are valuable resources at work to ensure the images are of the quality required for a perfect UI/UX.
Repetitive is the key term that stood out for us when we did our research. Anything repetitive can be automated. Better still, we could improve the output with Machine Learning. ImageEdit.ai was born of this idea to eliminate repetition. With ImageEdit.ai:
- Image editors can focus on work which uses their creative skills
- Images can be edited faster and can be scaled up & down on demand – acquiring computers is much easier than acquiring talent
- Editing errors are minimized & quality of images improve with the self-learning ability of Artificial Intelligence
- Use marketing resources for productive work than the quality check of edited images
- All this leads to faster product onboard – more sales
- Lesser cost of editing images – computers need power & not wages to outperform.
Pretty much a win-win situation for all stakeholders involved in the production of images!
How ImageEdit use AI to edit images?
To use Artificial Intelligence in editing images, we broke down the approach of editing images into 6 main categories: Clipping of the product from the background, adding or replacing the background, hair masking (like a removal of stray hair), garment & model retouch and introducing or rectifying shadows.
Our engineers worked on each category and created algorithms to automate & learn from every image edited. We call it Parallel Image Processing. Few hundred thousand images were processed through these algorithms till we achieved a desired quality of output.
While at it, brilliant minds from leading fashion institutes worked on each image to provide relevant feedback which was again fed into the system. This quality check process happened at three different stages on each image and observations were catalogued to make what we call a look-book – a personalized quality checklist for every client.
Below is a video on how every image flows through our algorithms:
So, What do you think about Image Editing using Artificial Intelligence? Share with us your thoughts!
