PDMS

Machine Learning

Mobile Application

Web Application

Procurement and Demand Management System

BUSINESS REQUIREMENT

A leading online market store for fish and livestock produce wanted a system to optimize the way procurement and demand of stock is managed. The fishery business currently involves a plethora of agents who supply the resource to major warehouses for distribution to households. The proposed system should be able to define demands, select agents, provide shipment details, track every shipment, keep track of the inventory and shipment details.

CHALLENGES
  • The integration of the Fish Identification tool into the Android app.
  • The lack of a dataset to train the Fish Identification tool.
SOLUTIONS
  • The super admin and admin functionalities are handled by a web dashboard, which allows them to add different users to the application along with adding warehouses. Super admin can allocate different users to a warehouse/many warehouses. Admin can handle the activities in their assigned warehouse/s and the different users like agents, supervisors, etc.
  • Supervisors and agents are provided with an android application to manage the warehouse, stocks, orders. Admin or supervisor can update demands of their warehouse daily/weekly or for a prolonged period, which are notified to the agents for fulfillment. The updates regarding an accepted demand can be tracked by the admin or supervisor.
  • The application also has a fish identification module that helps the agents and supervisors to identify the fish species while purchasing. This module was built using machine learning which fetches the name of fish while an image of the same is captured by the user.
  • The fish identification module was developed and integrated into the android app to help agents and supervisors to identify the fish while buying. The lack of dataset and model was tackled using Transfer Learning & Data Augmentation.
  • The tool was trained using TensorFlow & Object Detection API and was integrated into the app using TensorFlow Lite. The SSD Mobile Net V2 model was used along with hyperparameter tuning to improve the tool’s accuracy in predicting the fish.
IMPACT
  • The system has helped in updating requirements and tracking shipments using a web application and an android app.
  • The middlemen required from collecting fish from the harbor to a household were reduced thereby bringing down the resource cost.
  • The procurement of fish from agents to the warehouse is tracked and documented using this system.
  • The Fish Identification tool helps the supervisor and agent to identify the fish quickly and complete their duties in time.
KEY TECHNOLOGIES
  • JAVA
  • React JS
  • Mongo DB
  • Android
  • Tensor Flow
  • TensorFlow Lite
  • Dropwizard

More Projects

Category :