Keras implementation of ViT (Vision Transformer)

PyTorch and TensorFlow/Keras image models with automatic weight conversions and equal API/implementations - Vision Transformer (ViT), ResNetV2, EfficientNetV2, NeRF, SegFormer, MixTransformer, (planned...) DeepLabV3+, ConvNeXtV2, YOLO, etc.

A deepfake detection system combining Xception and Vision Transformer (ViT) models to identify AI-generated images and videos. Trained on the DFDC dataset, this project leverages TensorFlow/Keras for robust detection of facial manipulation artifacts.

基于Keras实现ViT模型，通过预训练权重在cifar-10数据集进行fine tuning，对图像分类准确率测试。

No description available

Keras implementation of Depth-Anything vits-model, inspired by Depth-Anything-ONNX architecture.

Vision Transformer (ViT) Tensorflow/Keras

Keras v3 (all backends) reimplementation of ViT models

An implementation of Vision Transformer (ViT) for classification of CIFAR 100 in Keras.

No description available

A reproducible cross‑framework study comparing CNN and CNN‑ViT hybrid architectures for EuroSAT satellite crop classification using aligned Keras and PyTorch implementations.

A complete deep learning pipeline for binary classification of satellite imagery into agricultural and non-agricultural land using CNNs, Vision Transformers, and CNN-ViT hybrid architectures — implemented in both Keras/TensorFlow and PyTorch.

Deep Learning project classifying satellite imagery into land-use categories using CNNs (ResNet50) and Vision Transformers (ViT). Developed in PyTorch and Keras, this model enables large-scale environmental monitoring and provides the framework for ground-level precision agriculture tasks.

Network Intrusion Detection System using Vision Transformer achieving 98.75% accuracy. The ViT model analyzes network traffic patterns to detect anomalies and malicious activity with high precision. Built with TensorFlow/Keras, it includes performance metrics and visualization tools for effective cybersecurity monitoring.

This TensorFlow/Keras model detects Alzheimer’s disease using MRI images. It uses transfer learning with a pre-trained MobileNetV2 as the base network and augments it with a custom Vision Transformer (ViT) head. Training is performed in two phases: initial training with a frozen base model and a fine-tuning phase where certain layers are unfrozen.

Vit実装(Keras)

ViT image classification using keras

No description available

exercising Swin-VIT

No description available

I have built a VIT classifier from scratch using TensorFlow and keras.

No description available

This project implements a Vision Transformer (ViT) architecture completely from scratch using TensorFlow / Keras — without using any pretrained ViT model.

Visual Transformer ViT model used in agricultural application

No description available

Basic Vision Transformer implementation from Scratch and Fine-tuning pretrained ViT using Tensorflow 2.0 Keras.

I have trained VIT transformer on set of images having binary class [0,1] using keras.

Vision-Transformer (ViT) is a state-of-the-art deep learning architecture for image recognition tasks. This repository contains an implementation of ViT in TensorFlow/Keras.

No description available

A ViT-CNN Hybrid Transformer trained on CIFAR-100 Dataset for categorical classification by using the Tensorflow and Keras frameworks.