Google Cloud Platform > Case Studies > autoRetouch: Revolutionizing Image Editing for Online Retailers with Google Cloud

autoRetouch: Revolutionizing Image Editing for Online Retailers with Google Cloud

Google Cloud Platform Logo
Technology Category
  • Analytics & Modeling - Computer Vision Software
  • Analytics & Modeling - Machine Learning
Applicable Industries
  • Education
  • Equipment & Machinery
Applicable Functions
  • Logistics & Transportation
  • Product Research & Development
Use Cases
  • Clinical Image Analysis
  • Predictive Maintenance
Services
  • Cloud Planning, Design & Implementation Services
  • Training
About The Customer

autoRetouch serves fashion image producers around the world, from retailers to editing services. These clients require high-quality images of their products for their online stores. The images often need edits such as background removal and skin retouching to ensure they accurately reflect the product being purchased. autoRetouch's clients are based all around the world, requiring the platform to be available around the clock. The platform's ability to scale from processing a few images to hundreds of thousands in seconds is crucial for these clients. The cost-effectiveness of the platform is also a significant factor, especially in the era of COVID-19, as companies shift their business online amid social distancing challenges.

The Challenge

autoRetouch, a Germany-based company, provides an automated image processing platform for fashion image producers worldwide. The shift to e-commerce has increased the demand for high-quality product images, which often require edits such as background removal and skin retouching. For retailers with hundreds of product lines, manually editing each image to perfection is a time-consuming and costly process. autoRetouch aimed to revolutionize this process by making it more time- and cost-effective. However, to train and run the models for its custom-developed machine learning algorithms, autoRetouch required powerful CPU and GPU processors that need to be available on demand. The platform had to be available around the clock and scale from processing a few images to hundreds of thousands, in seconds.

The Solution

autoRetouch decided to partner with Google Cloud to overcome its challenges. The company used an open source Google TensorFlow project on semantic image segmentation called DeepLab as a starting point for its models. After a year of hard work and multiple iterations, autoRetouch achieved a Mean Intersection over Union (MIoU) score of 99.7% in benchmarking categories, indicating a very high level of accuracy. To access enough computational power to run these models cost-effectively, autoRetouch turned to Google Cloud. The company uses AI Platform to host its models and run online predictions using Google Kubernetes Engine, scaling CPU-based models automatically. It also uses Cloud Storage buckets for both its training data and output data. To train its power-intensive machine learning models, the company uses GPUs on Compute Engine.

Operational Impact
  • autoRetouch's partnership with Google Cloud has not only streamlined its operations but also empowered its developers with more time to work on new features by reducing operational tasks to a minimum. The company runs a continuous deployment CI/CD pipeline, allowing it to roll out fixes and new features quickly. This has resulted in faster development and eliminated the need for dedicated operations personnel. The transparent billing on Google Cloud has also provided clear insight into the company's spending, helping it understand exactly what it's spending on. Furthermore, autoRetouch's ability to easily scale to meet the requests of retailers has supported business continuity for its clients during the COVID-19 pandemic, helping them shift their business online and reduce repetitive tasks across the image editing supply chain.

Quantitative Benefit
  • Up to 90% lower image retouching costs for end users through ML model training and processing on high-performance GPUs

  • Quick delivery of final images to end users by keeping processing times low whether one image is uploaded or thousands

  • Controls costs for the platform by providing easy scaling when computational load isn’t needed and per-second billing

Case Study missing?

Start adding your own!

Register with your work email and create a new case study profile for your business.

Add New Record

Related Case Studies.

Contact us

Let's talk!
* Required
* Required
* Required
* Invalid email address
By submitting this form, you agree that IoT ONE may contact you with insights and marketing messaging.
No thanks, I don't want to receive any marketing emails from IoT ONE.
Submit

Thank you for your message!
We will contact you soon.