Deep Learning Course Project (BLG – ITU)

This repository was created as part of a Deep Learning course project at
Istanbul Technical University (ITU).

---

Paper Studied

PnP Inversion: Boosting Diffusion-based Editing with 3 Lines of Code
International Conference on Learning Representations (ICLR), 2024

---

Project Scope

• Re-implementation of the proposed method using the authors’ publicly available code
• Qualitative image editing experiments on multiple examples
• Mini-benchmark evaluation on a subset of PIE-Bench
• Analysis of failure cases, with a focus on large pose changes
• Exploration of a potential improvement: mask-based localized diffusion and prompt scheduling
• Critical discussion and analysis of results, even when improvements are limited

---

Academic Use & Disclaimer

This work was conducted solely for academic and educational purposes as part of a university course project.

This repository is not an official implementation of the paper.
All original ideas, methods, and base code are credited to the paper authors.
Modifications, experiments, and analyses were performed by the student for learning and evaluation purposes.

---

Code Base

The project is based on the public implementation released by the paper authors.
The codebase was adapted, restructured, and extended for course-related experiments and analysis.

---

Citation

The original paper and codebase are cited below.

@article{ju2023direct,
  title={PnP Inversion: Boosting Diffusion-based Editing with 3 Lines of Code},
  author={Ju, Xuan and Zeng, Ailing and Bian, Yuxuan and Liu, Shaoteng and Xu, Qiang},
  journal={International Conference on Learning Representations (ICLR)},
  year={2024}
}