ChromaVision Models
This public repository contains the three model weight files used by the ChromaVision photo restoration pipeline. Public availability does not by itself grant reuse or redistribution rights; review the file-level terms in LICENSES.md.
Model files
| File | Pipeline stage | Purpose |
|---|---|---|
restoration_r34.pt |
Restoration | Restores degraded photographs. |
colorization_r34.pt |
Colorization | Adds color to grayscale photographs. |
RealESRGAN_x4plus.pth |
Enhancement | Upscales and enhances image detail with Real-ESRGAN. |
The files are separate checkpoints and are not merged. ChromaVision expects these filenames when loading the corresponding stages.
Project and training notebook
- ChromaVision source code: https://github.com/Pratyaksh0x1/ChromaVision
- Kaggle notebook: https://www.kaggle.com/code/pratyakshtomar0x1/chromavision
The notebook link is provided as project context and is not a guarantee that the training process or results have been independently reproduced.
Dataset references
The project owner reports the following datasets as training-data sources for the ChromaVision models. These are links only; this repository does not contain copies of the datasets. Access may require a Kaggle account and acceptance of the dataset or competition terms.
Colorization model
Restoration model
Dataset references do not grant permission to redistribute the datasets. Please consult each source's terms before downloading or using its data.
Licensing
The files have different origins and may have different terms. The Hub
repository-level license is marked other; this is not a single license for
all files. Read LICENSES.md before reusing or redistributing any
checkpoint. No license for the ChromaVision restoration or colorization
checkpoints is asserted here.
Limitations
Image restoration, colorization, and super-resolution can produce plausible details or colors that were not present in the original image. Review generated results before relying on them, especially for evidentiary, medical, or other high-stakes use. No evaluation metrics or performance guarantees are provided in this repository.