PROJECT TITLE :

Do-It-Yourself Recommender System: Reusing and Recycling With Blockchain and Deep Learning

ABSTRACT:

Waste is increasing at an exponential rate as a direct result of aggressive urbanization, which also contributes to the overall size of the human population. If we are able to reuse them, then this problem, which appears to be challenging, can be brought under control. In order to address this issue, the work that we do involves the design of a system that is oriented toward machine learning and blockchain technology. This system is capable of identifying waste objects and products and providing the user with multiple 'Do-It-Yourself' (DIY) ideas that can be reused or recycled. Every transaction that takes place on a blockchain is recorded in the network's shared ledger, which enables verifiability of transactions and facilitates improved decision-making. For the purpose of this research, a Deep Neural Network (DNN) that is capable of object recognition and was trained on approximately 11700 images using the ResNet50 architecture was developed. The training accuracy was 94%. To verify the do-it-yourself projects that have been recommended by members of the blockchain network, we use the Hyperledger Fabric (HF) blockchain platform to deploy several smart contracts. HF is a decentralized ledger technology platform that initializes and manages the ledger state by executing the deployed smart contracts within a secured Docker container. Our recommendation system operates on a web scraping script that was written in Python, and the entire model is distributed on a web platform that is built with the Flask framework. On a desktop computer outfitted with an Ubuntu 18.04 64-bit operating system, 16 gigabytes of random access memory (RAM), an Intel Core i7 processor with 8 cores, and the Python 3.6 package, retrieving do-it-yourself project ideas using web-scraping takes nearly one second. In addition, we use the hyperledger caliper benchmark to evaluate the performance of blockchain-based smart contracts in terms of their latencies and throughputs. To the best of our knowledge, this is the first piece of work that, for the purpose of a do-it-yourself recommender system, integrates blockchain technology and deep learning.