If you use this dataset in your research, please cite this publication:
Ying Chen, Darui Jin, Bin Guo, Xiangzhi Bai*. "Attention-Assisted Adversarial Model for Cerebrovascular Segmentation in 3D TOF-MRA Volumes." IEEE Transactions on Medical Imaging. 2022, 41(12): 3520-3532.

This dataset consists 1050 Thermal InfraRed images and pixel annotations of 8 categories in urban scenes. The categories and cooresponding labels are:
road: 0
sidewalk: 1
Pedestrian: 2
rider: 3
vehicle: 4
building: 5
vegetation: 6
sky: 7
background: 8
Images are given in \Img8bit and annotation files are given in \gtFine
The dataset is divided into train/val/test folders. Numbers of the images in the three sets are 50/50/950.

image

annotation (visulized)

If you use this dataset in your research, please cite this publication:
Junzhang Chen, Zichao Liu, Darui Jin, Yuanyuan Wang, Fan Yang, Xiangzhi Bai*, Light Transport Induced Domain Adaptation for Semantic Segmentation in Thermal Infrared Urban Scenes. IEEE Transactions on Intelligent Transportation Systems. 2022, 23(12), 23194-23211.

IRDST dataset consists of 142,727 real and simulation frames (40,650 real frames in 85 scenes and 102,077 simulation frames in 317 scenes). Each frame has labels of three types from fine to coarse: pixel-level mask, bounding box, central pixel. Pixel-level labels are given by labeling the targets in every pixel. Bounding-box labels are given by locating two-pixel away from border of pixel-level labels. Central pixel labels are given by locating the centroid of pixel-level labels.

Real data are given in \real and simulation data are given in \simulation
Images are given in \images
Pixel-level annotation files are given in \masks
Bounding box annotation files are given in \boxes
Central pixel annotation files are given in \center

image

Pixel-level annotation

If you use this dataset in your research, please cite this publication:
Heng Sun, Junxiang Bai, Fan Yang, Xiangzhi Bai*. Receptive-field and Direction Induced Attention Network for Infrared Dim Small Target Detection with a Large-scale Dataset IRDST. IEEE Transactions on Geoscience and Remote Sensing. 2023, 61:1-13.

BVTAMOS is a multi-centered dataset which includes 110 contrast enhanced CT scans of the abdomen from three public datasets (BTCV, TCIA and VISCERAL).
The accurate voxel-level manual segmentation of each volume is provided, which includes 14 main abdominal organs: spleen, right kidney, left kidney, gallbladder, esophagus, liver, stomach, aorta, inferior vena cava, portal vein and splenic vein, pancreas, right adrenal gland, left adrenal gland and duodenum.

If you use this dataset in your research, please cite this publication:
Hongjian Gao, Mengyao Lyu, Xinyue Zhao, Fan Yang. Xiangzhi Bai*, Contour-aware network with class-wise convolutions for 3D abdominal multi-organ segmentation, Medical Image Analysis. 2023, 102838.

TVSS is a thermal video semantic segmentation dataset, which consists of 1695 thermal videos in road scenes with 50850 frames in total. Each video is manually annotated with 17 categories at the frame rate of 1fps, including road, sidewalk, person, rider, passenger car, commercial vehicle, tricycle, two-wheeler, building, guard rail, bridge, pole, traffic sign, traffic light, vegetable, terrain, sky.

If you use this dataset in your research, please cite this publication:
Yu Zheng, Fugen Zhou, Shangying Liang, Wentao Song, Xiangzhi Bai*, Semantic Segmentation in Thermal Videos: A New Benchmark and Multi-Granularity Contrastive Learning based Framework. IEEE Transactions on Intelligent Transportation Systems, Accepted.

This contains the full version of the dataset (27,458 sequences with 411,870 frames) utilized in our paper "Neutralizing the impact of atmospheric turbulence on complex scene imaging via deep learning". Three main types of data are covered, which include algorithm simulated data, physical simulated data and real-world data. Specifically, the algorithm/physical simulated sequences are given with reference without turbulence distortion..

If you use this dataset in your research, please cite this publication:
Darui Jin, Ying Chen, Yi Lu, Junzhang Chen, Peng Wang, Zichao Liu, Sheng Guo and Xiangzhi Bai*, Neutralizing the impact of atmospheric turbulence on complex scene imaging via deep learning. Nature Machine Intelligence, 2021, volume 3, pages 876–884.

This dataset consists 5180 Thermal InfraRed images and pixel annotations of blind road. The categories and cooresponding labels are:
Background: 0
Blind road: 1
Images are given in \images and annotation files are given in \labels
The dataset is divided into train/val/test folders. Numbers of the images in the three sets are 2500/500/2180.

If you use this dataset in your research, please cite this publication:
Junzhang Chen, Xiangzhi Bai*, Atmospheric Transmission and Thermal Inertia Induced Blind Road Segmentation with a Large-Scale Dataset TBRSD. ICCV 2023.

This dataset contains atmospheric turbulence distorted infrared videos, clear videos, and corresponding atmospheric turbulence strength fields. The dataset provides 5,702 sets of simulation data, along with one real world captured turbulence distorted infrared video.

If you use this dataset in your research, please cite this publication:
Yadong Wang, Darui Jin, Junzhang Chen, Xiangzhi Bai*. “Revelation of hidden 2D atmospheric turbulence strength fields from turbulence effects in infrared imaging.” Nature Computational Science. 2023, 3: 687-699.

Yadong Wang, Xiangzhi Bai*. “An imaging-based approach to measure atmospheric turbulence.” Nature Computational Science. 2023, 3: 673-674.