A gap in online shopping and customer experience needs to be addressed, that despite creating ease for online buyers, a percentage of customers is still dissatisfied with online clothes shopping. Our aim was to conceive and configure a Proof of Concept (POC) that would suggest clothes to the buyer to try-on and then enable them to virtually try them on as an attempt to bridge this gap. The plan was to enable the online buyer to see if the clothes they were interested in, suit their body-type and skin-tone or not.
Team discussions enabled us to finalize the core functionality of the project that would be an application that could suggest a list of outfits to the user and then enable the user to virtually try them on. The user uploads their picture and lets the ML models work their magic.
We trained and tweaked with the most suitable ML models which would be able to work with disparities in skin tone, face shape, skin types and body types. GANs were used to give the user to choose the final image be either natural, anime or arcane inspired.
Our team thoroughly tested each process till we were able to eliminate the chances of sub-par results and until the final images were hyper-realistic which scaled the user experience.
The end result was a robust POC that suggests outfits to the user and then enables them to try those outfits on. It aligns and layers all the individually engineered components like face, hair and skin-tone on that of the fashion model to formulate a hyper-realistic image of the user in the clothes they wanted to try-on. A prototype was developed that gave the option to choose how the result image was displayed to the user. The final image could be Natural, Anime or even Hybrid- to scale the user experience.