There are crackerjack methods used to draw a clear picture of Machine Learning. One experiment was done, How-Old.net, the website was set up. The target was detecting the face, age, and gender from the face. The Face Analysis API feature was taken up for building this site. When the site was exhibited at the BUILD Conference, the application sprouted heat in the world. 85 million users adding up 500 million images for testing. The same thing stimulated the Media.

Coca Cola’s crusade to do something interesting on their 100-year anniversary absolutely nailed it to the market. The upcoming product of the company was Coca-Cola Contour Bottle. The brand decided to go for a grand fun way and unique celebration with its trusted Customers.  Coca-Cola decided to Partner with Universal McCann, the media Agency for honoring this breakthrough. The machine learning’s role and approach in this scenario were outstanding. The task was to detect the contour glass bottle in the photos. Coke had made an announcement to its fans, to come and join Coca-Cola in its remarkable celebration by clicking a selfie along with their new bottle of Coke and uploading a selfie on the site How-Old.net. The site was designed and aimed at detecting the face and presenting an age for the person and the Bottle as well.

There was one small hurdle which the Machine Learning needed to Cover up. The problem was that if the picture had the main focus on the human image and the bottle wasn’t having the complete focus then the Model had issues in detecting the Bottle properly.

The Model could detect the image only if the bottle was in better or complete focus.

Image Source: SPIE Digital Library

Machine Learning tried going two distinct ways:

The problem was resolved by combining two variant Machine Learning Algorithms.

1. Interest-Point Machine Algorithm: This helped in locating the logo on the Coca-Cola Bottle from wherever it was in the photograph submitted by the customer.
2. Deep Neural Network: This helped in locating if the Bottle was present in the submitted photograph, in the surrounding area of the logo.

Team Coca-Cola and Universal McCann, before implementing this Model, opted to go for a fair testing. This was done in between the users and with the Customers help. Users were asked to check the accuracy of their submission. The users can give their opinion on this by checking if the recognition went successfully or not. The users were asked to give a feedback on their experience in the testing Application created for their feedback.

There were several add-ins and cut-offs in the testing of this process.  With the help of the user’s feedback, focusing clearly on the accuracy and neatness of the performance of the model. The Model was approved to be introduced on the site of How-Old.net. It wasn’t a major task, just a single API call to a rest API operationalized on the Azure.

The users can still upload their high-resolution selfie of the Coca-Cola bottle with a visible logo on the site of How-Old.net. By this, they still get to unlock the Coca-Cola Experience and can see what happens after doing this. 

This was the finest strategy where The Team Coca-Cola can see how much their products are being Pampered and Microsoft can figure out how well their analytics is working and getting along with people’s experience.

Such Out of the Box methods of Machine Learning Models Congratulate the product and its Brand image.

There have been many such Innovations