| license: apache-2.0 | |
| library_name: onnx | |
| tags: | |
| - onnx | |
| - image-segmentation | |
| - image-colorization | |
| # Free tool models | |
| Browser ONNX weights for local image tools. Each file keeps its original Apache License 2.0 and the upstream attribution in that folder's `README.md`. | |
| ## image-colorizer/ddcolor-tiny-256-fp16.onnx | |
| DDColor-Tiny ([piddnad/DDColor](https://github.com/piddnad/DDColor)), exported to ONNX and converted to float16. About 108 MB. | |
| - Input `input`: float32 `[1, 3, 256, 256]`, the picture resized to 256×256, reduced to its CIE L channel and converted back to RGB (`a = b = 0`), values 0–1. | |
| - Output `output`: float32 `[1, 2, 256, 256]`, predicted CIE a and b channels. | |
| ## background-remover/isnet-general-use-fp16.onnx | |
| ISNet `isnet-general-use` ([xuebinqin/DIS](https://github.com/xuebinqin/DIS)), the ONNX build distributed by [rembg](https://github.com/danielgatis/rembg), converted to float16. About 86 MB. | |
| - Input `input_image`: float32 `[1, 3, 1024, 1024]`, RGB resized to 1024×1024, scaled to `x / 255 - 0.5`. | |
| - Output `output_image`: float32 `[1, 1, 1024, 1024]`, a saliency map. Min-max normalize it to get the alpha mask. | |
| sha256: | |
| - `ddcolor-tiny-256-fp16.onnx` `c50ff07e9c02285208bac88130dce46c3eb0961464f01e4597fe8f663761f8c7` | |
| - `isnet-general-use-fp16.onnx` `3fd7a4f645a1d330cb275611196bbb3d6687935bc429cf7bc21bfb095236333e` | |