Shivanshu Siyanwal

Quantum image classification

Quantum Image Representation Neural Networks

Quantum neural networks for binary image classification built on four quantum image encodings: NASS, QBIR, FTQR and FRQCI.

Published · The Journal of Supercomputing 2026 QML / Quantum image encoding / Qiskit

Interactive abstract

Play with the idea

This interactive diagram needs JavaScript. The schematic and write-up below explain the same idea.

Layman view

In plain words

Draft — to be verified. Written in plain language from the paper and project notes on this site.

A quantum computer cannot look at a picture directly: every pixel first has to be written into a quantum state, and there are many ways to do that. This study compared four of these quantum image encodings (NASS, QBIR, FTQR and FRQCI) to see which ones let a quantum neural network learn well.

Each encoding was placed in the same quantum neural network and trained on the same small images, from 2×2 pixels upwards, many times over with different random starting points. Besides accuracy, the study measured how large the resulting circuits are and whether training gets stuck.

The headline is that the encoding matters a great deal. Among the fully quantum models, QBIR gave the best balance of accuracy, noise robustness and circuit size, while FTQR and FRQCI stayed close to chance, a sign of barren plateaus, where the training signal fades away.

Expert view

The work, step by step

  1. 01

    Problem

    Quantum image classification depends heavily on how classical pixel data is encoded into quantum states, but encodings are often chosen by convention rather than by measured trade-off.

  2. 02

    Approach

    Four quantum image representations — Normal Arbitrary Superposition State (NASS), Quantum Block Image Representation (QBIR), Fourier-based Threshold Quantum Representation (FTQR) and Flexible Representation for Quantum Color Images (FRQCI) — are compared within one variational QNN framework. The main protocol runs the four encodings under a ring-entangled ansatz from 2×2 to 8×8; an ablation repeats the comparison with a hardware-efficient two-design ansatz and a strongly entangling ansatz, at all four resolutions on both datasets. Resource cost is measured as circuit depth, gate count and qubit count, and trainability through per-epoch gradient variance, L2 gradient norm, mean signed gradient and parameter-change magnitude. Every result is a mean ± σ over 20 random seeds.

  3. 03

    Results

    HCQNN was the strongest overall, with test accuracy up to 0.9756 ± 0.0076 (2×2, noiseless, two-design). Among the fully quantum models, QBIR gave the best balance of accuracy, noise robustness and depth: 0.9396 ± 0.0140 (2×2, noiseless, two-design), and the strongest fully quantum encoding at every resolution under Heron r3 noise. NASS was competitive at low resolution (up to 0.8928 ± 0.0196) but fell to 0.691 at 4×4 under the ring protocol. FTQR and FRQCI did not converge reliably, with test accuracy near chance (≈ 0.50–0.61), consistent with barren plateaus. At 16×16, QBIR needs depth 5,612, 9,701 gates and 16 qubits; FRQCI needs depth 125,299, 164,353 gates and 11 qubits.

  4. 04

    Status

    Published in The Journal of Supercomputing 82(8):495 (2026).

Schematic

How it works

103 104 105 8 10 12 14 16 18 Qubit count Circuit depth (log scale) 16×16 images QBIR 16 qubits · depth 5,612 FTQR 12 qubits · depth 43,273 NASS 9 qubits · depth 50,436 FRQCI 11 qubits · depth 125,299
Circuit resources of four quantum image encodings for 16×16 images: qubit count against circuit depth on a log scale. Data from Tables 1–2 of the J. Supercomputing paper.
Data (Tables 1–2)
Circuit resources at 16×16
EncodingQubitsDepthTotal gates
QBIR165,6129,701
FTQR1243,27359,683
NASS950,43672,713
FRQCI11125,299164,353

Contribution

My role

I conceived this study: the question of how the choice of quantum image encoding shapes the trainability, noise robustness and cost of variational QNNs. I designed the experimental protocol, selected and implemented the four encodings (NASS, QBIR, FTQR, FRQCI) in a single QNN framework, and carried out the comparative analysis, the visualisations and the organisation of results. During revision, in response to referee and editor feedback from The Journal of Supercomputing, I also carried out the state-of-the-art comparison, the gradient (trainability) analysis and the mathematical modelling.

  1. Study designMe (my part)
  2. Four encodings in one QNN frameworkMe (my part)
  3. Comparative analysisMe (my part)
  4. VisualisationMe (my part)
  5. Revision analyses (gradients, comparison, modelling)Me (my part)
My partCollaborators

Publication

The paper

2026 · Journal article

Quantum image representation-based quantum neural networks for binary classification

A. Tiwari*, S. Siyanwal*, S. Sakhuja, S. Kumar, P. Tyagi

The Journal of Supercomputing, 82(8), 495, 2026

doi:10.1007/s11227-026-08621-3

Abstract

A controlled comparison of four quantum image representations (NASS, QBIR, FTQR, FRQCI) inside one variational quantum neural network framework for image classification on near-term (NISQ) devices, tested on binary MNIST (0 vs 1) and a three-class FashionMNIST subset at resolutions from 2×2 to 16×16, under both noiseless simulation and a calibrated noise model of the IBM Heron r3 processor. Two reference models, a quanvolutional network (QCNN) and a hybrid CNN–QNN (HCQNN), place the encodings against established baselines.

* Equal contribution.

More threads