© 2019 Hey Machine Learning

Automatic Water Order

Problem

In companies where there is no office manager, the order is made by another person designated as responsible for the delivery of water. But because of the more important work cases for this period, he may forget to do it.

It is also very difficult to calculate the frequency of water consumption, as it is constantly changing due to the emergence of new employees, the time of year, a meeting with partners or internal events. The bottom line: there is no drinking water in the company for some time.

Solution

We created a system with Computer Vision that determines the number of full and empty bottles, and using Natural Language Processing, automatically calls the water delivery service and places an order.

 

The bottle recognition network is based on YOLO v3 and has been trained in 42 images. For speech processing, we used the Google Cloud Speech-to-Text cloud solution, which can extract meaning and analyze context from words. Dialogflow service was used to process the phrase and answer. To voice the decision was made using the service Google Cloud Text-to-Speech.

‚Äč

Among the technologies used are Python, YOLO, TensorFlow, Dialogflow, NLTK.

Result

Testimonials

“Solved the problem! Great! Very useful) Can we do this for office supplies? Often faced with the fact that the paper is over”

Anna

“I have 30 employees and in addition to a working business, there are 2 more startups from different areas, which take an unreal amount of time and attention. As a result, I physically can't even go to the kitchen once a week and count bottles. From the side it seems that there is nothing like taking 5 seconds, but in fact this is a huge problem, because all such trifles have to be kept in mind, and for owners it is extremely important to minimize the garbage in their heads. I used the Water Order system - life has become easier. Thanks”

Alex

“You are cool :)”

Denis

See more our works