There are two data files. Weibo.json is Micro-blog content, and relationship.json is Micro-blog forwarding relationship. 


To download the dataset click the link provided.  To unzip the file, double-click the zipped folder to open it. Then, drag or copy the item from the zipped folder to a new location.


The data set contains electrical and mechanical signals from experiments on three-phase induction motors. The experimental tests were carried out for different mechanical loads on the induction motor axis and different severities of broken bar defects in the motor rotor, including data regarding the rotor without defects. Ten repetitions were performed for each experimental condition.


The bench of experiments is on the premises of the School of Engineeringof São Carlos (EESC) of the University of São Paulo (USP), Brazil, more specifically in theLaboratory of Intelligent Automation of Processes and Systems (LAIPS) and Laboratory ofIntelligent Control of Electrical Machines (LACIME).

The three-phase induction motor is a model of the W22 standard line from manufacturer WEG, 1 cv, 220V / 380V, 3.02A / 1.75A, 4 poles, 60 Hz, with a nominal torque of 4.1 N.m and nominal speed of 1715 rpm. The rotor is a squirrel cage type made up of 34 bars. It is driven by means of a control panel that allows the selection of the type of drive, star or triangle, and the type of supply, direct mains voltage or via a three-phase inverter.

The rotary torque wrench used in the research is the Transtec model MT-103, with a maximum rotation of 2000 rpm, based on Wheatstone bridge technology and with a sensitivity of 2 mV / V. Its main function is to allow visualization of the torque present in the shaft, which will be varied simulating various operating conditions of the induction motor.

Manual adjustment of the resistant torque is done by varying the field winding voltage of the direct current generator. Therefore, to reduce the magnitude of the grid voltage, a 1800W single-phase voltage variation is used by Variac, and to convert the alternating voltage to continuous, a single-phase rectifier is used which feeds the field winding.

The vibration sensors used were Vibrocontrol uniaxial accelerometers, model PU 2001, with sensitivity of 10 mV / mm / s, frequency range 5 to 2000 Hz and stainless-steel housing, which provides the integrated acceleration signal over time. , ie provides the measure of vibration velocity. In total five accelerometers were used simultaneously, located non-drive end side motor, drive end side motor, housing, in the axial direction of the motor, and on the support desk. Therefore, these monitoring points allow the measurement of axial, tangential and radial velocity.

The currents were measured using alternating current probes, which correspond to precision meters, with a capacity of up to 50 A RMS, with an output voltage of 10 mV / A, corresponding to the Yokogawa model 96033. The voltages were measured directly at the MIT terminals using oscilloscope voltage tips also from the manufacturer Yokogawa.

To simulate the failure of broken bars in the squirrel cage rotor of the three-phase induction motor it was necessary to drill the rotor. Drilling was carried out by means of a bench drill mounted with a 6 mm diameter drill to ensure that the diameter of the hole exceeds the width of a rotor bar, with the tip centered at half the longitudinal length of the rotor.

Since in a real situation the breaking rotor bars are usually adjacent to the first broken bar, 4 rotors were tested, the first with one broken bar, the second with two adjacent broken bars, and so on to the rotor containing four adjacent bars. broken . It is worth mentioning that the hub depth of all tested rotors was the same, corresponding to 20 mm.

Thus, a rotor without a hole was tested first, that is, a healthy rotor, and then it was successively replaced in order to obtain a database of monitored variables.

Experiments were carried out using the bench mentioned above for the construction of the database. Tests were carried out on healthy motors and motors with defects in direct start with balanced three-phase supply voltage and 60 Hz frequency.

For the preparation of a reliable database, enabling future work were applied 0.5nm shipments, 1,0Nm, 1,5Nm, 2,0Nm, 2,5Nm, 3,0Nm, 3,5Nm, and 4.0Nm to the axis of the three-phase induction motor. For each loading condition of the motor shaft, ten repetitions were performed.

In this way, using the data acquisition system, for each experiment of each loading, the following variables were acquired:

·         voltages in phases A, B, and C;

·         currents in phases A, B, and C;

·         mechanical vibration speeds tangential in the housing, tangential in the base, axial on the driven side, radial on the driven side, and radial on the non-drive side.

This experimental process was performed for the detection and diagnosis of failures for healthy engines and engines with rotors containing 1, 2, 3, and 4 bars broken adjacent.

The database is organized as a structure of the Matlab application. The “struct_rs_R1” structure presents the experimental data referring to the defectless induction motor, “struct_r1b_R1” referring to the rotor with one broken bar, “struct_r2b_R1” referring to the rotor with two broken bars, “struct_r3b_R1” referring to the rotor with three broken bars and “Struct_r4b_R1” for the rotor with four broken bars.

When loading the files containing the experimental data for each structure in the Matlab application, it will be possible to view the experimental data for each of the mechanical loads imposed on the motor shaft. Then, it will be possible to observe the experimental data for each monitored variable.


Drone technology is one of the largest tackled fields in today’s world, as it can range from pure enjoyment of drone racing to medical use and fighting crime. Several teams are interested in developing improved human machine interfaces for operating drones. This dataset is a collection of different motion primitives commanded using a PS3 joystick to control an Ardrone on Gazebo. This has been conducted using ROS Melodic on Ubuntu 18.04.


To access the non-parsed dataset, which is a .YAML file, you can use any text editor program (i.e. Notepad, TextEdit…). In each of those files, you can see the sequence of inputs by the joystick which is updating every few milliseconds.


To access the parsed dataset, which is a .CSV file, you can use any spreadsheet program (i.e. Microsoft Excel, Google docs, Numbers…). In each of those files, there are 8 columns. First 3 columns are for x-y-z velocities, respectively. The 4th column represents the time (in seconds). The 5th, 6th, 7th, and 8th columns are for whether the button is pressed or not (1 for pressed, 0 for not pressed), and they are X, circle, square, and triangle respectively.


These last decades, Earth Observation brought quantities of new perspectives from geosciences to human activity monitoring. As more data became available, artificial intelligence techniques led to very successful results for understanding remote sensing data. Moreover, various acquisition techniques such as Synthetic Aperture Radar (SAR) can also be used for problems that could not be tackled only through optical images. This is the case for weather-related disasters such as floods or hurricanes, which are generally associated with large clouds cover.


The dataset is composed of 336 sequences corresponding to areas in West and South-East Africa, Middle-East, and Australia. Each time series is located in a given folder named with the sequence ID (0001... 0336).

Two json files, S1list.json and S2list.json are provided to describe respectively the Sentinel-1 and Sentinel-2 images.The keys are the total number of images in the sequence, the folder name, the geography of the observed area, and the description of each image in the series. The SAR images description contains also the URLs to download the images.Each image is described by its acquisition date, its label (FLOODING: boolean), a boolean (FULL-DATA-COVERAGE: boolean) indicating if the area is fully or partially imaged, and the file prefix. For SAR images the orbit (ASCENDING or DESCENDING) is also indicated.

The Sentinel-2 images were obtained from the Mediaeval 2019 Multimedia Satellite Task [1] and are provided with Level 2A atmospheric correction. For one acquisition, there are 12 single-channel raster images provided corresponding to the different spectral bands.

The Sentinel-1 images were added to the dataset. The images are provided with radiometric calibration and range doppler terrain correction based on the SRTM digital elevation model. For one acquisition, two raster images are available corresponding to the polarimetry channels VV and VH.

The original dataset was split into 267 sequences for the train and 67 sequences for the test. Here all sequences are in the same folder.


To use this dataset please cite the following papers:

Flood Detection in Time Series of Optical and SAR Images, C. Rambour,N. Audebert,E. Koeniguer,B. Le Saux,  and M. Datcu, ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2020, 1343--1346

The Multimedia Satellite Task at MediaEval2019, Bischke, B., Helber, P., Schulze, C., Srinivasan, V., Dengel, A.,Borth, D., 2019, In Proc. of the MediaEval 2019 Workshop


This dataset contains modified Copernicus Sentinel data [2018-2019], processed by ESA.

[1] The Multimedia Satellite Task at MediaEval2019, Bischke, B., Helber, P., Schulze, C., Srinivasan, V., Dengel, A.,Borth, D., 2019, In Proc. of the MediaEval 2019 Workshop


We propose a driver pattern dataset consists of 51 features extracted from CAN (Controller Area Network) of Hyundai YF Sonata while four drivers drove city roads of Seoul, Republic of Korea. Under the belief that different driving patterns implicitly exist at CAN data, we collected CAN diagnosis data from four drivers in pursuit of research on driver identification, driver profiling, and abnormal driving behavior detection. Four drivers are named A, B, C, and D.



The dataset contains 51 features extracted from CAN along with numerous trips performed by four drivers. The four drivers drove along city roads of Seoul, the Republic of Korea. The recorded 51 features can be employed for driver identification, driver profiling, abnormal driving pattern identification, and any related tasks. Please check the abstract for a more detailed description.

CSV Files

Directory A, B, C and D contains .csv files of CAN data. Each .csv file represents a trip.


The names of 51 features are described in the features.pkl file. Please check the file for detailed information.


Park, Kyung Ho, and Huy Kang Kim. "This Car is Mine!: Automobile Theft Countermeasure Leveraging Driver Identification with Generative Adversarial Networks." arXiv preprint arXiv:1911.09870 (2019).

Park, Kyung Ho, and Huy Kang Kim. "This Car is Mine!: Automobile Theft Countermeasure Leveraging Driver Identification with Generative Adversarial Networks.", ESCAR Asia (2019)



Classification and Regression folders each have a 'zz_published_results'folder. You can either generate your own results again using the Or you can move these folders'contents up to either the Classification or Regression folder and onlyrun the analysis code in sf_analysis.


Recently, the coronavirus pandemic has made the use of facial masks and respirators common, the former to reduce the likelihood of spreading saliva droplets and the latter as Personal Protective Equipment (PPE). As a result, this caused problems for the existing face detection algorithms. For this reason, and for the implementation of other more sophisticated systems, able to recognize the type of facial mask or respirator and to react given this information, we created the Facial Masks and Respirators Database (FMR-DB).


For reasons related to the copyright of the images, we cannot publish the entire database here. If you are a student, a professor, or a researcher and you want to use it for research purposes, send an email to attaching the license, duly completed, which you can find here on IEEE DataPort.



This multispectral remote sensing image data contained pixels of size (1024 x 1024) for the region around Kolkata city in India and was obtained with LISS-III sensor. There are four spectral bands, i.e., two from visible spectrum (green and red) and two from the infrared spectrum (near-infrared and shortwave infrared). The spatial resolution and spectral variation over the wavelength are 23.5m and 0.52 - 1.7 μm, respectively.