ImageSieve vision model attribution and conversion notice

CLIP ViT-L/14 DataComp.XL
Publisher: LAION / DataComp
https://huggingface.co/laion/CLIP-ViT-L-14-DataComp.XL-s13B-b90K
Reviewed model-card revision: 84c9828e63dc9a9351d1fe637c346d4c1c4db341
License declared by that model card: MIT.
The model card describes any deployed use case as outside its recommended
scope without task-specific testing. ImageSieve performs local image search
and organization, not surveillance or facial recognition.

SigLIP 2 SO400M/14 (384)
Publisher: Google
https://huggingface.co/google/siglip2-so400m-patch14-384
Reviewed model-card revision: e8e487298228002f3d8a82e0cd5c8ea9c567f57f
License declared by that model card: Apache-2.0.

Modified distribution: ImageSieve distributes separate Core ML image and
text encoders converted from the upstream PyTorch models, with fixed input
shapes and a SentencePiece tokenizer. The conversion implementation is
swift-server/siglip2_to_coreml.py in the ImageSieve source repository.
Conversion changes the storage and inference representation; it does not
imply endorsement by Google, LAION or DataComp. Exact historical weight
and converter revisions are not established by the model-card revisions.
The model archive SHA-256 values in VisionModelDescriptor.swift identify
the distributed bytes. No claim of new training or fine-tuning is made.

The complete license texts accompany this notice. This notice does not
replace any rights or conditions in the upstream licenses.