| raon-vision-encoder |
| Copyright 2024-2026 Raon Vision Team |
|
|
| This product includes software derived from the following projects: |
|
|
| =============================================================================== |
| OpenCLIP |
| https://github.com/mlfoundations/open_clip |
| Licensed under the MIT License (see LICENSES/MIT-OpenCLIP.txt) |
|
|
| Copyright (c) 2012-2021 Gabriel Ilharco, Mitchell Wortsman, |
| Nicholas Carlini, Rohan Taori, Achal Dave, Vaishaal Shankar, |
| John Miller, Hongseok Namkoong, Hannaneh Hajishirzi, Ali Farhadi, |
| Ludwig Schmidt |
|
|
| Used in: model/ and train/ packages (LocCa, CLIP, loss, factory, |
| transformer, data pipeline, training loop, etc.) |
|
|
| =============================================================================== |
| OpenAI CLIP |
| https://github.com/openai/CLIP |
| Licensed under the MIT License (see LICENSES/MIT-OpenAI-CLIP.txt) |
|
|
| Copyright (c) 2021 OpenAI |
|
|
| Used in: model/tokenizer.py, model/bpe_simple_vocab_16e6.txt.gz |
|
|
| =============================================================================== |
| Meta Platforms, Inc. (MAE / MoCo v3) |
| Licensed under the MIT License via OpenCLIP |
|
|
| Copyright (c) Meta Platforms, Inc. and affiliates |
|
|
| Used in: model/pos_embed.py (sincos position embedding utilities) |
|
|
| =============================================================================== |
| timm (pytorch-image-models) |
| https://github.com/huggingface/pytorch-image-models |
| Licensed under the Apache License 2.0 |
|
|
| Copyright (c) Ross Wightman |
|
|
| Used in: model/transform.py (ResizeKeepRatio) |
|
|
| =============================================================================== |
| References |
|
|
| The following papers informed the design and implementation of features |
| in this software. Code was independently implemented unless noted above. |
|
|
| - CoCa: Yu et al., "CoCa: Contrastive Captioners are Image-Text Foundation Models", 2022 |
| - SigLIP: Zhai et al., "Sigmoid Loss for Language Image Pre-Training", 2023 |
| - SigLIP2: Tschannen et al., "SigLIP 2: Multilingual Vision-Language Encoders", 2025 |
| - DINO: Caron et al., "Emerging Properties in Self-Supervised Vision Transformers", 2021 |
| - DINOv2: Oquab et al., "DINOv2: Learning Robust Visual Features without Supervision", 2024 |
| - SILC: Naeem et al., "SILC: Improving Vision Language Pretraining with Self-Distillation", 2023 |
| - TIPS: Huang et al., "TIPS: Text-Image Pretraining with Spatial Awareness", 2024 |
| - Koleo: Sablayrolles et al., "Spreading vectors for similarity search", ICLR 2019 |
| - Gram Anchoring: Simeoni et al., "DINOv3", 2025 (independently implemented) |
| - NaFlex: from SigLIP2 / PaLI (independently implemented in PyTorch) |
|
|
