ps-clip-labeler

Fast Selection Labeler artifacts for the DepthField Photo desktop app: a CLIP ViT-B/32 vision encoder plus a precomputed text-embedding table over a curated vocabulary. Used to identify the subject of a selection (auto-naming layers, Auto Enhance categorization).

Contents & attribution

Derived from OpenAI CLIP (model openai/clip-vit-base-patch32).

• Original author: OpenAI
• Source: https://github.com/openai/CLIP
• License: MIT (full text below)

clip-vision-encoder.onnx

• Modifications: exported to ONNX from the CLIP vision tower with the visual-projection head fused, so the encoder output already lives in the joint 512-dim embedding space.

clip-text-embeddings.bin, clip-vocab.json

• Derived works: text embeddings precomputed by running a curated noun vocabulary through CLIP's text tower and L2-normalizing. The vocabulary list is original to DepthField Photo.

License

MIT License

Copyright (c) 2021 OpenAI

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.