@camstack/addon-ai 0.4.124 → 0.4.125
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/THIRD_PARTY_MODELS.md +8 -0
- package/dist/addon.js +640 -124
- package/dist/addon.mjs +640 -124
- package/package.json +1 -1
package/THIRD_PARTY_MODELS.md
CHANGED
|
@@ -61,6 +61,12 @@ under [Non-commercial models](#non-commercial-models).
|
|
|
61
61
|
https://github.com/ShiqiYu/libfacedetection.train.
|
|
62
62
|
Trained on WIDER FACE, whose images are CC BY-NC-ND.
|
|
63
63
|
- **AuraFace v1** — fal.ai, Apache-2.0. https://huggingface.co/fal/AuraFace-v1
|
|
64
|
+
- **SigLIP 2** — © Google LLC, Apache-2.0 (`Apache-2.0.txt`).
|
|
65
|
+
https://huggingface.co/google/siglip2-base-patch16-224 ·
|
|
66
|
+
https://github.com/google-research/big_vision. CamStack's SigLIP 2 files are
|
|
67
|
+
modified: the vision input scaling `2x - 1` is baked into the graph, the text
|
|
68
|
+
graph slices its input to the model's 64 positions, the tokenizer lower-cases,
|
|
69
|
+
and the towers are converted to ONNX, OpenVINO IR and CoreML FP16.
|
|
64
70
|
- **EasyOCR** — JaidedAI, Apache-2.0. https://github.com/JaidedAI/EasyOCR
|
|
65
71
|
- **U²-Net** — Xuebin Qin et al., Apache-2.0. https://github.com/xuebinqin/U-2-Net
|
|
66
72
|
- **YAMNet** — © Google LLC, Apache-2.0.
|
|
@@ -202,6 +208,8 @@ commercial use**:
|
|
|
202
208
|
| --- | --- | --- | --- | --- | --- |
|
|
203
209
|
| `mobileclip-s1`, `mobileclip-s2` (image) | Apple `ml-mobileclip` | MIT / **Apple Machine Learning Research Model License** | per Apple | Model Derivative: converted to OpenVINO IR and CoreML FP16 | Mirror `clip/mobileclip-s1`, `clip/mobileclip-s2` |
|
|
204
210
|
| `mobileclip-s1-text`, `mobileclip-s2-text` | Apple `ml-mobileclip`, with the CLIP tokenizer | MIT / **Apple Machine Learning Research Model License** | per Apple | Model Derivative: quantised to INT8 ONNX and converted to OpenVINO IR | Mirror `clip/mobileclip-s1`, `clip/mobileclip-s2` |
|
|
211
|
+
| `siglip2-b16-224` (image) | Google SigLIP 2 base patch16-224 (`google/siglip2-base-patch16-224`) | Apache-2.0 / Apache-2.0 | WebLI (Google internal, not released) | input scaling `y = 2x - 1` baked ahead of the patch embedding; converted to OpenVINO IR and CoreML FP16 (an FP32 ONNX is hosted as the verification reference) (`scripts/build-siglip2-model.py`) | Mirror `clip/siglip2` |
|
|
212
|
+
| `siglip2-b16-224-text` | Google SigLIP 2 base patch16-224, with its Gemma tokenizer | Apache-2.0 / Apache-2.0 | WebLI (Google internal, not released) | input sliced to the model's 64 positions in the graph; ONNX weights FP16, OpenVINO IR and CoreML FP16; a `Lowercase` normalizer prepended to the tokenizer | Mirror `clip/siglip2` |
|
|
205
213
|
|
|
206
214
|
### Segmentation
|
|
207
215
|
|