Add Zenodo DOI badge and citation entry
Browse files
README.md
CHANGED
|
@@ -18,6 +18,10 @@ language:
|
|
| 18 |
|
| 19 |
# Masafee LoRA
|
| 20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
Stable Diffusion LoRA for the original character **Masafee** — a red panda from a personal [LINE sticker set](https://store.line.me/stickershop/product/19794926/ja).
|
| 22 |
Trained entirely on a single consumer GPU (NVIDIA GeForce RTX 3060, 12GB) with [kohya-ss/sd-scripts](https://github.com/kohya-ss/sd-scripts).
|
| 23 |
|
|
@@ -68,6 +72,25 @@ The full experiment writeup (Japanese / English, two-column PDF), all training s
|
|
| 68 |
2. **Epoch ~30 is the sweet spot** — extending to 40–60 epochs produced visible overfitting without improving character fidelity.
|
| 69 |
3. **Output quality is bounded by training image count** — with only 13 stickers, certain poses and lighting conditions remain out of distribution.
|
| 70 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
## License
|
| 72 |
|
| 73 |
- **LoRA weights (this file):** released for **personal and research use only**. The weights encode a specific copyrighted character.
|
|
@@ -76,10 +99,6 @@ The full experiment writeup (Japanese / English, two-column PDF), all training s
|
|
| 76 |
|
| 77 |
By downloading this model you agree to use it only for personal experimentation or academic research.
|
| 78 |
|
| 79 |
-
## Citation
|
| 80 |
-
|
| 81 |
-
If you reference this work, please link the GitHub repository above. A Zenodo DOI is planned.
|
| 82 |
-
|
| 83 |
---
|
| 84 |
|
| 85 |
-
Made with 🐾 by [masafykun](https://github.com/masafykun) · [masafy.org](https://masafy.org/)
|
|
|
|
| 18 |
|
| 19 |
# Masafee LoRA
|
| 20 |
|
| 21 |
+
[](https://github.com/masafykun/masafee-lora)
|
| 22 |
+
[](https://doi.org/10.5281/zenodo.20397946)
|
| 23 |
+
[](https://orcid.org/0009-0000-7977-2756)
|
| 24 |
+
|
| 25 |
Stable Diffusion LoRA for the original character **Masafee** — a red panda from a personal [LINE sticker set](https://store.line.me/stickershop/product/19794926/ja).
|
| 26 |
Trained entirely on a single consumer GPU (NVIDIA GeForce RTX 3060, 12GB) with [kohya-ss/sd-scripts](https://github.com/kohya-ss/sd-scripts).
|
| 27 |
|
|
|
|
| 72 |
2. **Epoch ~30 is the sweet spot** — extending to 40–60 epochs produced visible overfitting without improving character fidelity.
|
| 73 |
3. **Output quality is bounded by training image count** — with only 13 stickers, certain poses and lighting conditions remain out of distribution.
|
| 74 |
|
| 75 |
+
## Citation
|
| 76 |
+
|
| 77 |
+
If you use this model in academic work, please cite it via the Zenodo archive:
|
| 78 |
+
|
| 79 |
+
```bibtex
|
| 80 |
+
@software{suzuki_masafee_lora_2026,
|
| 81 |
+
author = {Suzuki, Masato},
|
| 82 |
+
title = {{Masafee LoRA: Single-GPU LoRA Fine-tuning of an
|
| 83 |
+
Original Character on Stable Diffusion}},
|
| 84 |
+
year = {2026},
|
| 85 |
+
version = {v1.0.0},
|
| 86 |
+
doi = {10.5281/zenodo.20397946},
|
| 87 |
+
url = {https://doi.org/10.5281/zenodo.20397946},
|
| 88 |
+
orcid = {0009-0000-7977-2756}
|
| 89 |
+
}
|
| 90 |
+
```
|
| 91 |
+
|
| 92 |
+
**Persistent archive (Zenodo DOI):** [10.5281/zenodo.20397946](https://doi.org/10.5281/zenodo.20397946)
|
| 93 |
+
|
| 94 |
## License
|
| 95 |
|
| 96 |
- **LoRA weights (this file):** released for **personal and research use only**. The weights encode a specific copyrighted character.
|
|
|
|
| 99 |
|
| 100 |
By downloading this model you agree to use it only for personal experimentation or academic research.
|
| 101 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 102 |
---
|
| 103 |
|
| 104 |
+
Made with 🐾 by [masafykun](https://github.com/masafykun) · [masafy.org](https://masafy.org/) · [ORCID](https://orcid.org/0009-0000-7977-2756)
|