# Vision Transformer

Based on the OpenCLIP-ViT-L-14 model, the Test-Time Register technology is introduced to improve the model's interpretability and the performance of downstream tasks.

A Vision Transformer-based binary classification model specifically designed for detecting deepfake images, with an accuracy rate of 99.2%

This paper proposes a method to reinterpret Vision Transformer (ViT) as an image segmentation model, demonstrating ViT's potential in image segmentation tasks.

This paper proposes a method to reinterpret Vision Transformer (ViT) as an image segmentation model, revealing ViT's potential in image segmentation tasks.

This paper proposes an image segmentation model based on Vision Transformer (ViT), revealing the potential of ViT in image segmentation tasks.

This paper proposes a method to reinterpret Vision Transformer (ViT) as an image segmentation model, demonstrating ViT's potential in image segmentation tasks.

This model reveals the potential of Vision Transformer (ViT) in image segmentation tasks by adapting its architecture specifically for segmentation.

The model proposed in this paper reveals the potential of Vision Transformer (ViT) in image segmentation tasks.

This paper proposes a novel perspective by treating Vision Transformer (ViT) as an image segmentation model and explores its potential in image segmentation tasks.

This model is developed based on the paper 'Your ViT is Actually an Image Segmentation Model' and is a Vision Transformer model for image segmentation tasks.

This model reveals the potential of Vision Transformer (ViT) in image segmentation tasks by transforming ViT into an efficient image segmentation model through specific methods.

This model reveals the potential of Vision Transformer (ViT) in image segmentation tasks by adapting its architecture for segmentation purposes.

This paper proposes a method to reinterpret Vision Transformer (ViT) as an image segmentation model, demonstrating ViT's potential in image segmentation tasks.

By rethinking the architecture of Vision Transformer (ViT), this model demonstrates its potential in image segmentation tasks.

A smart TV gesture recognition model based on the Vision Transformer architecture, capable of accurately classifying 9 common gestures.

A Vision Transformer (ViT)-based image classification model for distinguishing AI-generated from human-created images, achieving 98% accuracy.

ViTPose++ is a vision Transformer-based foundational model for human pose estimation, achieving an outstanding performance of 81.1 AP on the MS COCO keypoint test set.

ViTPose++ is a vision Transformer-based foundation model for human pose estimation, achieving an outstanding performance of 81.1 AP on the MS COCO keypoint test set.

ViTPose++ is a vision Transformer-based human pose estimation model, achieving outstanding performance of 81.1 AP on the MS COCO keypoint detection benchmark.

ViTPose is a vision Transformer-based human pose estimation model that achieves an outstanding performance of 81.1 AP on the MS COCO keypoint detection benchmark with a simple design.

ViTPose is a human pose estimation model based on Vision Transformer, achieving outstanding performance on benchmarks like MS COCO through simple architectural design.

A vision Transformer-based human pose estimation model achieving an outstanding performance of 81.1 AP on the MS COCO keypoint test set

ViTPose is a human pose estimation model based on Vision Transformer, achieving 81.1 AP accuracy on the MS COCO keypoint test set, with advantages such as model simplicity, scalable size, and flexible training.

AIM-v2 is an image encoder model based on the timm library, with a 3B parameter scale, suitable for image feature extraction tasks.

AIM-v2 is an image encoder model based on the timm library, suitable for image feature extraction tasks.

This is a vision transformer model based on DINOv2, which improves the attention mechanism by adding register tokens for unsupervised image feature extraction.

This model is a facial expression classifier fine-tuned on the FER2013 dataset based on Google's ViT base model, capable of classifying images into four facial expressions.

This model is used to detect whether an image is real or AI-generated, employing the Vision Transformer (ViT) architecture for precise classification.

ViTPose is a baseline model for human pose estimation based on plain vision transformers, achieving high-performance keypoint detection with a simple architecture

A dual-framework compatible vision model trained on the MetaCLIP-2.5B dataset, supporting both OpenCLIP and timm frameworks

A dual-framework compatible vision model trained on the MetaCLIP-400M dataset, supporting both OpenCLIP and timm frameworks

An end-to-end structured optical character recognition system specifically designed for Arabic, converting book page images into structured text (Markdown format)

An image classification model based on Google's Vision Transformer (ViT) architecture, specifically designed to identify five hairstyle types (curly, dreadlocks, twists, straight, wavy) from facial images with 93% accuracy.

Sapiens is a series of vision transformer models pre-trained on 300 million human images at 1024x1024 resolution, focusing on human-centric vision tasks.

Sapiens is a Vision Transformer model pre-trained on 300 million high-resolution human images, specializing in human-centric vision tasks

Sapiens is a vision Transformer model pre-trained on 300 million 1024×1024 resolution human images, supporting high-resolution inference and real-world scenario generalization.

Sapiens is a Vision Transformer model pre-trained on 300 million high-resolution human images, specializing in human-centric vision tasks.

Sapiens is a family of Vision Transformer models pre-trained on 300 million 1024x1024 resolution human images, specializing in human-centric vision tasks.

Sapiens is a Vision Transformer model pre-trained on 300 million human images, specializing in human-centric segmentation tasks and supporting 1K high-resolution inference.

Sapiens is a Vision Transformer model pre-trained on 300 million 1024×1024 resolution human images, excelling in human-centric vision tasks.