transunet

The Tensorflow, Keras implementation of U-net, V-net, U-net++, UNET 3+, Attention U-net, R2U-net, ResUnet-a, U^2-Net, TransUNET, and Swin-UNET with optional ImageNet-trained backbones.

The repository for paper AA-TransUNet: Attention Augmented TransUNet For Nowcasting Tasks.

Here are the codes for the "TransU-Net++: Rethinking attention gated TransU-Net for deforestation mapping" paper.

This research enhances early disease diagnosis by analyzing retinal blood vessels in fundus images using deep learning. It employs eight pre-trained CNN models and Explainable AI techniques.

Welcome, Awesome-segmentation-in-medical project. This repo is robust, reproducible, and fair benchmarking framework for breast ultrasound (BUS) image segmentation. Features 10+ models (UNet, UNet++, TransUnet, SwinUnet etc.), 5+ datasets, and strict data separation to prevent test set leakage.

An ML model with U-shaped architecture with ResNet50V2 and Vision Transformer based encoders

A hybrid CNN–Transformer framework for precise industrial surface defect detection and segmentation, integrating Vision Transformer (ViT) with convolutional modules to effectively capture both local texture details and global contextual features.

AI-powered maritime surveillance system that detects ships and oil spills from SAR satellite imagery in real time. Built with TransUNet, Deformable DETR, and a gigapixel image viewer for human-in-the-loop validation.

Projects based on self learning purpose.

SOTA 3D Medical Segmentation framework (UNET++, TransUNET , SwimTransUNET) in TensorFlow. Fully TPU-optimized with a 3-stage OHEM pipeline for the Medical Segmentation Decathlon (MSD).

Code for our submission to Solafune: Identifying Deforestation Drivers ML competition

MSCAF-TransUNet: Multi-Scale CNN Attention Fusion for Synapse segmentation. Current result: 76.61 Dice / 28.80 HD95.

Medical image segmentation using U-Net and TransUNet on the ISIC 2018 dataset. Evaluated CNN vs Transformer architectures and implemented progressive unfreezing with differential learning rates, achieving a Dice Similarity Coefficient of 0.9139 while training under 15 GB GPU memory constraints.

Official PyTorch code for a two-stage pancreas CT segmentation pipeline: Faster R-CNN localization + TransUNet with a Dice-Hausdorff loss.

Abdominal multi-organ segmentation workbench built with PyTorch, TransUNet and PySide6

6th AI SPARK Challenge (2024): wildfire segmentation from satellite images with TransUNet and Attention U-Net

Advanced AI/ML slide decks and literature reviews.

Semester 8 Data Science Project: Medical Image Segmentation with Vision Transformers

EBRT dose prediction: Geant4/OpenGATE Monte Carlo dataset + 3D TransUNet+FiLM surrogate for prostate VMAT (gamma 3%/3mm 76.3% vs TPS)