CaaStle Inc is a US-based technology company that offers a clothing rental service to its customers. Since its establishment in 2012, the company has partnered with top fashion brands and retailers to provide subscribers with a broad range of high-quality clothing items for a flat monthly fee.
Introduction
Background
Like many other companies, CaaStle Inc had to adapt to a new working environment due to the pandemic. One of the challenges that arose was the inability to take pictures of the garments displayed on the website. To address this issue, the company decided to use the images provided by the brand/supplier. However, these images presented challenges in terms of their dimensions and the placement of the models.
Figure 1 – Potential automated cropping issues
Methodology
To ensure that the models were not cropped at the head or feet, the company developed a custom solution using a public AI library. The algorithm was designed to detect the position of the model and crop the picture accordingly while centering the image on the model and avoiding cutting off any parts of the model.
Two possible issues either the supplied picture once resized was too large or too tall as seen on Figure 1.
Figure 2 – Smart cropping process
We had a couple of options on how to handle this situation
- Manual cropping
- “Dumb” automatic cropping
- Smart automatic cropping
- 3rd party tool
- In-house – “hack”
- In-house – fully fledge solution
In order to make a decision the following criteria were considered
- Time to market
- Engineering resources needed
- Pricing
- Human handling
- Capabilities of the tool
Based on the above the in-house hack solution was picked and was able to deliver great value to the company.
Results
The automated cropping of the images through our in-house “hack” saved countless human hours, making the process more efficient and cost-effective for the company.