MixMe photo-edit prototype models (Core ML)
Core ML conversions used by a photo-editing prototype (research spike, MIXME-408). Downloaded on demand by the app; not part of any shipping product.
| Model | Task | Size | Input | Output | Source & licence |
|---|---|---|---|---|---|
DDColorTiny.mlpackage |
B&W photo colorization | ~116 MB | 512×512 RGB image (gray rendered from L) | ab tensor (1×2×512×512) |
DDColor (ddcolor_paper_tiny), Apache-2.0 |
RealESRGANx2.mlpackage |
2× super-resolution | ~33 MB | 512×512 RGB tile | 1024×1024 RGB image | Real-ESRGAN (RealESRGAN_x2plus), BSD-3-Clause |
Both converted with coremltools (fp16 mlprogram, iOS 17 minimum target) from the authors' official released weights, and parity-checked against the PyTorch originals.
Credits: DDColor © 2023 piddnad et al. (ICCV 2023); Real-ESRGAN © 2021 Xintao Wang et al. (ICCVW 2021). See the linked repositories for the original licences.
- Downloads last month
- 78
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support