WIN1205-05 Josephine Management Co. Ltd.


 

Hi, my name is Tirth Vyas and I am 22 year old. I am from Gujarat, India and I am a Bachelor in Information Communication and Technology from Ahmedabad University. I am currently enrolled in Wireless Information Networking at Fleming College. My expertise is in Computer Vision and Networking.

I am Vishal Hegde from Bangalore, India. I have a bachelor degree in electronics and electrical engineering. My strength are integration of hardware and software. I am currently perusing Wireless information networking in Fleming college. My hobbies are to play football and watch TV series.

I completed my bachelor’s in Computer Science & Engineering. I had done project in PHP, Java and C++ during my last semester of Bachelors. I also have knowledge about C, HTML, SQL and Artificial Intelligence.

My name is Chintankumar Patel and I am 22 Years old. I am from Gujarat, India and I have completed by Bachelor of Application from Kadi Sarva Vidhyalaya. I am proficient in app development and software engineering.


Team Name:

Sharanga

Team Members Names:

Tirth Vyas, Vishal Hegde, Milan Patel, Chintankumar Patel

Program of Study:

Wireless Information Networking

Mentor Name:

Cathy Smits

Project Name:

WIN1205-05 Josephine Management Co. Ltd.

Sponsor Name:

Glenn Verkindt



Sponsor Organisation:

Josephine Management Co. Ltd.


Brief Description Of Project:

We created a concept of automated inventory management system that will detect the number of bottles and position of bottles in a wine-rack and display the result in a basic inventory management app to facilitate a better inventory and bookkeeping experience. This project is meant to be a side accessory to the Cellar Genius Rack which is being developed by Josephine Management Co. Ltd.

Project Objective:

  • The determination of an empty wine rack by using separate camera and integrating it with a mobile app. 
  • The design and develop a system that can recognize individual slots and determine if they empty or occupied by a bottles.
  • With the help of an application, we must be able to know which row-column is empty or full so that the client can easily stack up the wine bottles in the rack.
  • Further improve the speed of the algorithm.

Project Innovation:

  • Use of Computer Vision to extract data from the captured image of the rack and determine the occupied bottle slots.
  • Incorporation of Machine Learning to improve accuracy of the algorithm while detecting bottles in the wine rack.
  • JSON encoding and decoding for send data to the iOS App.
  • State based object detection for each slot to further improve the speed of the algorithm.

Challenges & Solutions:

  • We had not experience of iOS app development and computer vision at the start of the project, so we used various resources from the internet to learn and apply.
  • Entire body bottles was not visible since the camera’s position was on the top so we focused only on the caps of the bottles.
  • We were able to detect the bottle caps but we were not able to determine the position, so we divided each image of the rack into several parts and run the algorithm on each of them. We call these parts the areas of interest.
  • When we used Neural Network to recognize bottle there was lot of computational overhead, since there were 126 bottles in the rack it would take 5 to 7 mins to detect all the bottles. To counter act that we added an other algorithm that senses a state change of each slot. Therefore, we can use the neural network algorithm to detect all the bottles when the system is switched on and then we can just keep on detecting change in each slots, so it is much faster and it is night and day difference.

Project Results or Progress:

  • We successfully designed and tested an algorithm utilizing machine learning and image processing to determine if the bottle slots in a rack are occupied. We developed an iOS App which can fetch data from the system running the algorithm to display the results in an intuitive way.
  • The entire system is still at a concept stage and given more time it can be developed in a complete package that can be included in the box along with the Rack.

Lessons Learned:

  • How to train and apply neural networks to extract features form a frame and determine the presence of a specific object. Here we learned applications of python language.
  • We acquired networking skills by using compression and decompression of jSON data and transmitting it over the network.
  • We also got to learn iOS development in order to develop an basic app to display the results, we learned SwiftUI and Swift.

Short Video: