Quantum Threads

An end to end under water animal identification machine learning model using Quantum kernels along with Classical pretrained models. This project comprises a brief research on three popular classical models ResNet18 , ResNet50 ,InceptionV3 and its quantum counter parts Hybrid Quantum classical ResNet18 | ResNet50 | InceptionV3.

Base solution

• To provide a web app backboned by Hybrid quantum classical CNN models
• Hybrid Quantum ML model will use PQC as Quantum kernels along with pretrained classical models
• Whole model will be trained on UAD-2023 dataset containing >13k images for 23 classes
• Entire model is made learnable , quantum kernels will use classical gradient descent for optimization

What's Innovative ?

Distributed kernel processing

• Distributing all kernels across multiple cloud QPU's for parallelism

Adjoint differentiation

• Using reversible nature of quantum circuits to compute gradients , thus bringing time complexity from exponential to constant

Demo

Environment Variables

If you want to run inference on real IBM Quantum device you need to add IBM Quantum token to config.toml file

API_KEY

Run locally

Clone this repo

  git clone https://github.com/nagarajRPoojari/Quantum-Threads.git

install requirements

    pip install -r requirements.txt

start streamlit server

    streamlit run main.py

Authors

• @nagarajRPoojari
• @nithinmkannal
• @Varun-rocky

Tech Stack

• Tensorflow
• Pennylane
• Qiskit
• Amazon braket
• streamlit
• AWS

License

MIT