Clip Vit Base Patch16 Alternatives

A port of OpenAI's CLIP ViT-B/32 model by Xenova, optimized for execution in JavaScript environments via ONNX Runtime or WebNN. It enables zero-shot image classification, image-text similarity, and multimodal embeddings directly in the browser or on the server with Node.js. The model is widely used for client-side computer vision tasks.

clip-vit-large-patch14-336 is an open-source vision-language model by OpenAI that enables joint image and text embedding for tasks such as zero-shot classification and retrieval. It is widely used by AI researchers and developers for multimodal applications.

This is a Vision Transformer (ViT-B/16 at 224px) model from the timm library, loaded with OpenAI CLIP weights for image feature extraction. It is compatible with both timm and Transformers libraries and has been widely downloaded. The model produces embeddings suitable for similarity search, classification, or other computer vision downstream tasks.

This model is a Vision Transformer (ViT) base with patch size 16 trained using the CLIP objective on the LAION-400M dataset (epoch 31). It is part of the timm library and supports zero-shot image classification and image embedding generation. The model is distributed on Hugging Face for use with the OpenCLIP library.

chinese-clip-vit-base-patch16 is a Vision Transformer (ViT) based CLIP model trained by OFA-Sys for Chinese vision-language understanding. It supports zero-shot image classification, image-text retrieval, and related tasks using Chinese text. The model is available on Hugging Face with Transformers integration and is suitable for developers working with Chinese multimodal content.

This model is a Vision Transformer (ViT) trained with the CLIP objective on the LAION-400M dataset using the OpenCLIP library. It is hosted in the timm model hub and supports zero-shot image classification as well as image and text embedding generation for downstream computer vision tasks.

This is a Vision Transformer (ViT) model trained with the CLIP objective on the LAION-400M dataset. It enables zero-shot image classification and image-text similarity computation. The model is provided through the timm library and OpenCLIP on Hugging Face.

X-CLIP is a multimodal model developed by Microsoft that extends CLIP to the video domain. It learns joint embeddings between video clips and text, enabling zero-shot video classification, retrieval, and other video-language tasks. The base-patch32 variant is available on Hugging Face for integration into computer vision and multimedia applications.

Vit Base Patch14 Reg4 Dinov2.lvd142m is a pretrained model hosted on Hugging Face for image feature extraction. It belongs to the class of vision transformer models and was developed within the timm PyTorch Image Models library. The model carries an apache-2.0 license and is made available as an open artifact for integration into computer vision workflows. It can be loaded directly through the timm library by calling timm.create_model with the identifier "hf_hub:timm/vit_base_patch14_reg4_dinov2.lvd142m" and the pretrained flag set to true. The same checkpoint supports the Transformers library, where developers instantiate it via AutoModel.from_pretrained or through a pipeline configured for the image-feature-extraction task. These loading paths enable use in notebooks such as Google Colab or Kaggle as well as in local applications and inference providers. The model card references three arXiv papers that document the underlying techniques. Its primary output consists of image representations suitable for downstream tasks, though no additional training details or performance metrics appear on the page. The repository provides safetensors and PyTorch format files along with version history.

Vit Base Patch14 Dinov2.lvd142m is a pre-trained image feature extraction model hosted on Hugging Face. It belongs to the class of Vision Transformer models and was developed within the timm PyTorch Image Models library. The model carries the apache-2.0 license and draws on research described in arxiv papers 2304.07193 and 2010.11929. It supports direct use through the timm library by creating the model with a single line that loads pretrained weights from the Hugging Face hub. The same weights are also accessible via the Transformers library, where developers can instantiate it with AutoModel or employ the image-feature-extraction pipeline for straightforward inference. Model files are provided in both PyTorch and Safetensors formats. The entry appears under the Image Feature Extraction task tag. It is delivered exclusively as a downloadable model artifact rather than a standalone application, allowing integration into notebooks, local scripts, or inference services that support timm or Transformers. No pricing information is listed because the model is fully open source.

CLIP_VITB32_512 is a light image embedding model in EIDORA for fast visual grouping. It is described as a broadly useful visual model for image work, with an emphasis on first-pass grouping when semantic similarity matters more than exact object classification. Its listed uses center on fast first-pass visual grouping, image collections, and other EIDORA image workflows. The model card says it is not ideal for fine-grained visual similarity when a heavier model is acceptable, for text-only projects, or for specialized domains that need a domain-trained visual encoder. The page also identifies it as a feature extraction model and associates it with embeddings, image, and clip. Delivery is through ONNX, with onnxruntime named on the page. The model is tagged with openai/webimage-text and eidora-model-zoo, and it is labeled compute:light and modality:image. The compute tier note says it has a small download, low memory use, and faster CPU runtime, and it is described as running on most laptops. The license is MIT. The page also lists arxiv: 2103.00020 and shows the maker or host as EIDORA / eidora.org.

laion/CLIP-ViT-B-32-laion2B-s34B-b79K is an open-source vision-language model that generates embeddings for both images and text. It is widely used for zero-shot classification, search, and retrieval tasks, providing a robust foundation for multimodal AI applications. The model is suitable for researchers and developers building advanced AI systems.

This repository provides ONNX exports of the OpenCLIP ViT-B-32 model originally from OpenAI. It is specifically intended for use within Immich, a self-hosted photo and video library management application. The model enables image feature extraction and semantic search capabilities.

This model is a large Vision Transformer (ViT) trained with the SigLIP objective, made available through the timm library. It is designed for image feature extraction and can be used with both the timm and transformers libraries for downstream computer vision tasks.

This is a CLIP ViT-B/16 model trained on the DataComp XL dataset (s13B-b90K). It is designed for vision-language tasks including zero-shot image classification and retrieval. The model is available on Hugging Face and uses a standard CLIP architecture for aligning image and text embeddings in a shared space.

This is an Apple-developed variant of the CLIP model using a ViT-B/16 vision transformer backbone and a Data Filtering Network (DFN). It produces aligned embeddings for images and text, enabling zero-shot classification, retrieval, and other vision-language tasks. The model is available on Hugging Face for use with the Transformers library.

siglip-base-patch16-224 is an open-source vision-language model developed by Google for zero-shot image classification. It enables users to classify images without task-specific training by leveraging multimodal embeddings. Ideal for researchers and developers in computer vision and AI.

This LAION model combines a ViT-B/32 image encoder with an XLM-RoBERTa-base text encoder, trained on the LAION-5B dataset. It is part of the OpenCLIP ecosystem and supports cross-lingual image-text similarity, retrieval, and zero-shot classification. The model is distributed as open weights and integrates directly with the OpenCLIP library.

google/owlvit-base-patch32 is an open-vocabulary vision transformer model from Google Research that performs zero-shot object detection. It can detect arbitrary objects in images based on text queries without being limited to a fixed set of classes. The model is hosted on Hugging Face and integrates with the transformers library, making it useful for researchers and developers building flexible computer vision applications.

This is an open-weight CLIP model variant trained by LAION on the LAION-2B dataset. It maps images and text into a shared embedding space for tasks such as zero-shot image classification, image-text retrieval, and multimodal similarity scoring. The model is distributed on Hugging Face and can be used with the OpenCLIP or transformers libraries.

CLIP-ViT-B-16-laion2B-s34B-b88K is a vision transformer model trained with the CLIP objective on the large-scale LAION-2B dataset. It produces aligned embeddings for images and text that support zero-shot classification, retrieval, and similarity tasks. The model weights are publicly available on Hugging Face for research and commercial applications.