Research map / Image generation / Editing and style transfer
Hybrid Quantum-Classical Style Transfer (Student Abstract)
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes quantum style transfer (QST), replacing classical CNN down/upsampling in a style-transfer network with quantum convolutional layers (torchquantum simulation) and showing qualitative visualizations versus a classical CNN.
Problem
Style transfer with classical CNNs has limited representational capacity for separating content from style, and the authors aim to reduce inference time and complexity while maintaining performance.
Approach
Replace the classical convolutional down- and upsampling layers of a style-transfer network with quantum convolution layers (encoding, variational quantum circuit, decoding) in a hybrid quantum-classical architecture, using stride-based down/upsampling and residual connections.
Key results
- QCNN-based style transfer successfully transforms the style of input images under all tested style images, per visual inspection (Fig. 2).
- QCNN outputs differ visibly from classical CNN outputs, which the authors present as offering greater image diversity.
- No quantitative metrics for inference time, complexity, or performance are reported despite the abstract's claims.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace classical CNN with QCNN | QST · Style transfer visualization (Figure 2) | style transformation | Helped |
| Replace classical CNN with QCNN | QST · Style transfer visualization (Figure 2) | image diversity | Helped |
Surprising findings
- The abstract claims reduced inference time and complexity, but the body reports no timing or complexity measurements.
Limitations
- Evaluation is purely qualitative visualization with no quantitative metrics.
- Experiments run on quantum simulators (torchquantum), not real quantum hardware.
- Real-time processing and generalization to more domains remain open challenges.
Experimental setup
Metrics: qualitative visual comparison
Baselines: classical CNN convolution strategy
Builds on: QCNN torchquantum DCGAN-style transposed convolution architecture
Closest papers by meaning
- Quantum Visual Fields with Neural Amplitude Encoding (NeurIPS 2025 · similarity 0.82)
- Quantum-PEFT: Ultra parameter-efficient fine-tuning (ICLR 2025 · similarity 0.80)
- Structured Unitary Tensor Network Representations for Circuit-Efficient Quantum Data Encoding (NeurIPS 2026 · similarity 0.80)
- Characterizing Trainability of Instantaneous Quantum Polynomial Circuit Born Machine (NeurIPS 2026 · similarity 0.80)
- QuCOOP: A Versatile Framework for Solving Composite and Binary-Parametrised Problems on Quantum Annealers (CVPR 2025 · similarity 0.80)
- Quantum Doubly Stochastic Transformers (NeurIPS 2025 · similarity 0.79)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.