In this competition, you will have to develop an algorithm for automatic categorization of products by their name and attributes, even in conditions of incomplete marking.

The category system is arranged in the form of a hierarchical tree (up to 5 levels of nesting), and product data comes from many trading platforms, which creates a number of difficulties:

• Product marking is incomplete and imperfect.
• The attributes of the same products may differ on different platforms or be absent altogether.
• As the catalog expands, new platforms and categories may appear that did not exist before.

In this competition, we invite participants to try their hand at setting up a task as close to the real one as possible:

• The task is complicated by the fact that in the training sample, categories are marked only for half of the products, and in the test sample there will be categories for which there are no marked examples.
• At the same time, you are provided with a complete list of possible categories in advance - will you be able to take them all into account and accurately assign products to the right classes?
• LLMs (Large Language Models) look promising for such tasks – but will they really help? This is a great chance to test different approaches!

The cosmetic modification of hair is a very common procedure used to mask or cover evidence at a crime scene. Deoxyribonucleic acid (DNA) tests are expensive and require good-quality collection of samples and a database profile. To overcome these challenges, direct analysis was performed on a large set of hair strands collected from individuals, denoted original samples, and the data were compared with those of the same samples after cosmetic modification performed by bleaching the samples in the laboratory. A total of 127 samples were evaluated in this study using two analytical techniques, wavelength-dispersive X-ray fluorescence (WDXRF) and laser-induced breakdown spectroscopy (LIBS). Instead of testing many algorithms to develop classification models for the original and bleached samples, a recent method was applied that combines information from 17 classifiers. Data fusion was also evaluated to improve the accuracy of the classification model, which was higher than 99.2%, with no requirements to select eigenvectors or thresholds.

https://www.grss-ieee.org/wp-content/uploads/2022/01/DFC2022-IADF-example-training-set.jpg" alt="example">

BackgroundIn the last decade data fusion has become widespread in the field of metabolomics. Linear data fusion is performed most commonly. However, many data display non-linear parameter dependences. The linear methods are bound to fail in such situations. We used proton Nuclear Magnetic Resonance and Gas Chromatography-Mass Spectrometry, two well established techniques, to generate metabolic profiles of Cerebrospinal fluid of Multiple Sclerosis (MScl) individuals. These datasets represent non-linearly separable groups. Thus, to extract relevant information and to combine them a special framework for data fusion is required. MethodologyThe main aim is to demonstrate a novel approach for data fusion for classification; the approach is applied to metabolomics datasets coming from patients suffering from MScl at a different stage of the disease. The approach involves data fusion in kernel space and consists of four main steps. The first one is to extract the significant information per data source using Support Vector Machine Recursive Feature Elimination. This method allows one to select a set of relevant variables. In the next step the optimized kernel matrices are merged by linear combination. In step 3 the merged datasets are analyzed with a classification technique, namely Kernel Partial Least Square Discriminant Analysis. In the final step, the variables in kernel space are visualized and their significance established. ConclusionsWe find that fusion in kernel space allows for efficient and reliable discrimination of classes (MScl and early stage). This data fusion approach achieves better class prediction accuracy than analysis of individual datasets and the commonly used mid-level fusion. The prediction accuracy on an independent test set (8 samples) reaches 100%. Additionally, the classification model obtained on fused kernels is simpler in terms of complexity, i.e. just one latent variable was sufficient. Finally, visualization of variables importance in kernel space was achieved.

According to Cognitive Market Research, the global Finance Data Fusion market size was USD 11251.2 million in 2024. It will expand at a compound annual growth rate (CAGR) of 25.20% from 2024 to 2031.

North America held the major market share for more than 40% of the global revenue with a market size of USD 4500.48 million in 2024 and will grow at a compound annual growth rate (CAGR) of 23.4% from 2024 to 2031.
Europe accounted for a market share of over 30% of the global revenue with a market size of USD 3375.36 million.
Asia Pacific held a market share of around 23% of the global revenue with a market size of USD 2587.78 million in 2024 and will grow at a compound annual growth rate (CAGR) of 27.2% from 2024 to 2031.
Latin America had a market share of more than 5% of the global revenue with a market size of USD 562.56 million in 2024 and will grow at a compound annual growth rate (CAGR) of 24.6% from 2024 to 2031.
Middle East and Africa had a market share of around 2% of the global revenue and was estimated at a market size of USD 225.02 million in 2024 and will grow at a compound annual growth rate (CAGR) of 24.9% from 2024 to 2031.
The Large Enterprises segment had a larger market share in 2023

Market Dynamics of Finance Data Fusion Market

Key Drivers for Finance Data Fusion Market

Advancements in AI and Machine Learning

Advances in AI and ML in Finance Data Fusion are growth drivers. Technologically speaking, these advancements have enabled this industry to quickly analyze such enormous financial data to greater accuracy and, therefore allow different organizations to diagnose hidden areas or trends of financial opportunities. AI and machine learning algorithms can automate data processing, improve predictive analytics, and enable fast decision-making, which is critical in the high speed of the present-day financial world. As companies increasingly realize that such technologies will enable them to streamline their operations and provide the best customer experience, they are bound to invest in finance data fusion solutions. Rising demand not only propels market interest but also ignites innovation and sophisticated tools and applications for more innovative practices in the financial sector. For instance, in 2022, LexisNexis introduced an AI-powered risk assessment platform that employs machine learning to evaluate consumer data and detect potential fraud. This enables organizations to improve their compliance and risk management efforts.

Growing Adoption of Cloud Computing

The adoption of cloud computing is a key driver of growth in the Global Finance Data Fusion Market. Cloud computing provides various benefits to financial organizations, including scalability, flexibility, and cost-effectiveness. Financial firms that use cloud-based finance data fusion solutions can avoid investing in expensive on-premises infrastructure and scale their data fusion capabilities as needed. Cloud computing also gives financial firms access to a broader range of data and services. For example, cloud-based data fusion systems can combine information from social media, public data sources, and third-party data providers. This allows financial organizations to have a more thorough understanding of their consumers and the market. In addition, cloud computing allows financial organizations to experiment with novel data fusion technologies and applications. This helps them remain ahead of the competition and create unique products and services.

Restraint Factor for the Finance Data Fusion Market

High Implementation Costs act as a restraint to the market

High implementation costs remain one of the significant drawbacks against growth in the Finance Data Fusion Market. Spending money by organizations during the investment of the newest integration tools and technologies, coupled with the basic infrastructures required to support these entities, such as the purchase of software licenses, hardware, and maintenance over time, bears down on small to medium-sized companies. Apart from that, the number of trained personnel to handle and analyze the data will increase in cost. In other instances, some businesses will not adopt finance data fusion solutions, thereby limiting the potential growth and innovation within the market. This restriction slows down the collective adoption of necessary data-driven strategies in the financial sector.

Opportuni...

The dataset is captured over Samford Ecological Research Facility (SERF), which is located within the Samford valley in south east Queensland, Australia. The central point of the dataset is located at coordinates: 27.38572oS, 152.877098oE. The Vegetation Management Act 1999 protects the vegetation on this property as it provides a refuge to native flora and fauna that are under increasing pressure caused by urbanization.The hyperspectral image was acquired by the SPECIM AsiaEAGLE II sensor on the second of February, 2013. This sensor captures 252 spectral channels ranging from 400.7nm to 999.2nm. The last five channels, i.e., channels 248 to 252, are corrupted and can be excluded. The spatial resolution of the hyperspectral data was set to 1m.The airborne light detection and ranging (LiDAR) data were captured by the ALTM Leica ALS50-II sensor in 2009 composing of a total of 3716157 points in the study area: 2133050 for the first return points, 1213712 for the second return points, 345.736 for the third return points, and 23659 for the fourth return points.The average flight height was 1700 meters and the average point density is two points per square meter. The laser pulse wavelength is 1064nm with a repetition rate of 126 kHz, an average sample spacing of 0.8m and a footprint of 0.34m. The data were collected up to four returns per pulse and the intensity records were supplied on all pulse returns.The nominal vertical accuracy was ±0.15m at 1 sigma and the measured vertical accuracy was ±0.05m at 1 sigma. These values have been determined from check points contrived on an open clear ground. The measured horizontal accuracy was ± 0.31m at 1 sigma.The obtained ground LiDAR returns were interpolated and rasterized into a 1m×1m digital elevation model (DEM) provided by the LiDAR contractor, which was produced from the LiDAR ground points and interpolated coastal boundaries.The first returns of the airborne LiDAR sensor were utilized to produce the normalized digital surface model (nDSM) at 1m spatial resolution using Las2dem.The 1m spatial resolution intensity image was also produced using Las2dem. This software interpolated the points using triangulated irregular networks (TIN). Then, the TINs were rasterized into the nDSM and the intensity image with a pixel size of 1m. The intensity image with 1m spatial resolution was also produced using Las2dem.The LiDAR data were classified into ground" andnon-ground" by the data contractor using algorithms tailored especially for the project area. For the areas covered by dense vegetation, less laser pulse reaches the ground. Consequently, fewer ground points were available for DEM and nDSM surfaces interpolation in those areas. Therefore, the DEM and the nDSM tend to be less accurate in these areas.In order to use the datasets, please fulfill the following three requirements:

1) Giving an acknowledgement as follows:

The authors gratefully acknowledge TERN AusCover and Remote Sensing Centre, Department of Science, Information Technology, Innovation and the Arts, QLD for providing the hyperspectral and LiDAR data, respectively. Airborne lidar are from http://www.auscover.org.au/xwiki/bin/view/Product+pages/Airborne+LidarAirborne hyperspectral are from http://www.auscover.org.au/xwiki/bin/view/Product+pages/Airborne+Hyperspectral

2) Using the following license for LiDAR and hyperspectral data:

http://creativecommons.org/licenses/by/3.0/3) This dataset was made public by Dr. Pedram Ghamisi from German Aerospace Center (DLR) and Prof. Stuart Phinn from the University of Queensland. Please cite: In WORD:Pedram Ghamisi and Stuart Phinn, Fusion of LiDAR and Hyperspectral Data, Figshare, December 2015, https://dx.doi.org/10.6084/m9.figshare.2007723.v3In LaTex:@article{Ghamisi2015,author = "Pedram Ghamisi and Stuart Phinn",title = "{Fusion of LiDAR and Hyperspectral Data}",journal={Figshare},year = {2015},month = {12},url = "10.6084/m9.figshare.2007723.v3",

}

Sensor Fusion Platform Market Outlook

According to our latest research, the global Sensor Fusion Platform market size reached USD 7.8 billion in 2024, reflecting a robust expansion driven by technological advancements and surging demand across multiple industries. The market is expected to grow at a CAGR of 17.5% during the forecast period, with the market size projected to reach USD 36.5 billion by 2033. This remarkable growth is primarily fueled by the increasing adoption of connected devices, advancements in artificial intelligence, and the imperative need for real-time data processing in applications such as autonomous vehicles, smart consumer electronics, and industrial automation. As per our latest research, the rapid integration of sensor fusion platforms is fundamentally transforming data analytics and decision-making processes across sectors.

One of the primary growth factors for the sensor fusion platform market is the proliferation of smart devices and the Internet of Things (IoT). With the exponential rise in IoT-enabled applications, there is a growing requirement for platforms that can seamlessly combine data from multiple sensors to deliver accurate, real-time insights. Sensor fusion enables the integration of data from various sources—such as inertial, image, radar, and ultrasonic sensors—resulting in enhanced reliability and precision. This capability is particularly critical in applications like autonomous vehicles and wearable devices, where safety and user experience are paramount. Additionally, the increasing complexity of modern devices necessitates sophisticated data processing platforms, further fueling the demand for advanced sensor fusion solutions.

Technological advancements in artificial intelligence and machine learning are also acting as significant catalysts for market growth. Sensor fusion platforms are increasingly leveraging AI algorithms to process and interpret vast volumes of sensor data, enabling smarter and more adaptive systems. For example, in the automotive sector, AI-powered sensor fusion platforms are crucial for advanced driver-assistance systems (ADAS) and autonomous driving, allowing vehicles to make split-second decisions based on comprehensive environmental data. Furthermore, the integration of edge computing with sensor fusion platforms is empowering industries to perform real-time analytics at the source, reducing latency and enhancing operational efficiency. These innovations are expected to drive widespread adoption and open new opportunities across diverse sectors.

Another key growth driver is the expanding application scope of sensor fusion platforms beyond traditional industries. While automotive and consumer electronics have historically dominated the market, sectors such as healthcare, aerospace, defense, and industrial automation are now recognizing the value of sensor fusion. In healthcare, for instance, sensor fusion platforms are being used to monitor patient vitals and enhance diagnostic accuracy. In industrial settings, they enable predictive maintenance and process optimization by aggregating data from multiple sensors. The ability to deliver actionable insights from heterogeneous data sources is positioning sensor fusion platforms as indispensable tools for digital transformation initiatives across the board.

A critical component of this technological evolution is the Sensor Fusion Processor, which plays a pivotal role in the seamless integration of data from multiple sensors. This processor acts as the brain of the sensor fusion platform, efficiently managing and processing the vast amounts of data generated by various sensors. By utilizing advanced algorithms and high-speed processing capabilities, the Sensor Fusion Processor ensures that data is accurately combined and interpreted in real time. This capability is essential for applications such as autonomous vehicles, where quick and precise decision-making is crucial for safety and performance. The development of more sophisticated sensor fusion processors is expected to further enhance the capabilities of sensor fusion platforms, enabling them to support increasingly complex applications across different industries.

From a regional perspective, Asia Pacific is emerging as the largest and fastest-growing market for sensor fusion platforms, driven by the rapid adoption of advanced technologies

A new data set for automated driving, which is the subject matter of this paper, identifies and addresses a gap in existing similar perception data sets. While the most state-of-the-art perception data sets primarily focus on provision of various on-board sensor measurements along with the semantic information under various driving conditions, the provided information is often insufficient since the object list and position data provided include unknown and time-varying errors. The current paper and the associated data-set describes the first publicly available perception measurement data that include not only the on-board sensor information from camera, Lidar and radar with semantically classified objects, but also the high precision ground-truth position measurements enabled by the accurate RTK assisted GPS localization systems available on both the ego vehicle and the dynamic target objects. This paper provides insight on the capturing of the data, explicitly explaining the meta data structure and the content, as well as the potential application examples where it has been, and can potentially be, applied and implemented in relation to automated driving and environmental perception systems development, testing and validation.

The CARbon DAta MOdel fraMework (CARDAMOM; Bloom et al., 2015 in review) outputs are derived from a global 1-degree x 1-degree 2001-2010 model-data fusion (MDF) analysis. The datasets include allocation fractions (AF) residence times (RT), mean carbon pool stocks (CP) and fluxes (FL). A list of files and their contents is provided below. The Data Assimilation Linked Ecosystem Carbon model version 2 (DALEC2) and the Markov Chain Monte Carlo MDF algorithm are described by Bloom & Williams (2015); the fire module is described by Bloom et al., (2015; in review). Data constraints used in the CARDAMOM analysis consist of MODIS leaf area index (LAI), Harmonised World Soil Database (HWSD; Hiederer & Kochy, 2012) and tropical biomass (Saatchi et al., 2011). For each 1-degree x 1-degree gridcell, the metrics (e.g. mean, median, etc.) are based on 4000 DALEC2 model parameter samples unique to that grid-cell. We note that the full 2001-2010 CARDAMOM output amounts to roughly 10 TB in binary format. For the sake of brevity, we have limited the following datasets to the subset presented in Bloom et al., (2015, in review). Additional MDF outputs can be made available upon request. ##CONTACTS:## Anthony Bloom: abloom@jpl.nasa.gov Mathew Williams: mat.williams@ed.ac.uk Jeff Exbrayat: j.exbrayat@ed.ac.uk ##Datasets:## ###File name: Description ### ------------------------ CARDAMOM_2001_2010_AF_AUT.nc: GPP fraction autotrophically respired. CARDAMOM_2001_2010_AF_LAB.nc: GPP fraction allocated to labile C. CARDAMOM_2001_2010_AF_FOL.nc: GPP fraction allocated to foliar C. CARDAMOM_2001_2010_AF_ROO.nc: GPP fraction allocated to fine root C. CARDAMOM_2001_2010_AF_WOO.nc: GPP fraction allocated to wood C. CARDAMOM_2001_2010_RT_FOL.nc: Foliar C residence time CARDAMOM_2001_2010_RT_ROO.nc: Fine root C residence time CARDAMOM_2001_2010_RT_WOO.nc: Wood C residence time CARDAMOM_2001_2010_RT_LIT.nc: Litter C residence time CARDAMOM_2001_2010_RT_SOM.nc: Soil carbon residence time CARDAMOM_2001_2010_CP_LAB.nc: Mean 2001-2010 labile C CARDAMOM_2001_2010_CP_FOL.nc: Mean 2001-2010 foliar C CARDAMOM_2001_2010_CP_ROO.nc: Mean 2001-2010 fine root C CARDAMOM_2001_2010_CP_WOO.nc: Mean 2001-2010 woody C CARDAMOM_2001_2010_CP_LIT.nc: Mean 2001-2010 litter C CARDAMOM_2001_2010_CP_SOM.nc: Mean 2001-2010 soil C CARDAMOM_2001_2010_FL_GPP.nc: Gross primary production CARDAMOM_2001_2010_FL_NPP.nc: Net primary production CARDAMOM_2001_2010_FL_RAU.nc: Autotrophic respiration CARDAMOM_2001_2010_FL_RHE.nc: Heterotrophic respiration CARDAMOM_2001_2010_FL_FIR.nc: Fires CARDAMOM_2001_2010_FL_NEE.nc: Net ecosystem exchange CARDAMOM_2001_2010_FL_NCE.nc: Net carbon exchange CARDAMOM_2001_2010_LCMA.nc: Leaf mass per area CARDAMOM_2001_2010_NCE_monthly_mode.nc: Mode monthly NCE CARDAMOM_2001_2010_FIGURE_MAPS.nc: Datasets used to make figures 1-3. ###NOTES:### + AF*, CP* and LCMA netcdf (.nc) files: Lon, Lat and global 180x360 (LatxLon) datasets: (mean, median, st. dev, 5th, 25th, 75th, 95th %iles) + RT* files: (log-based mean, median, log-based st. dev, 5th, 25th, 75th, 95th %iles) + FL* files: Lon, Lat and global 180x360 (LatxLon) datasets: (mean, median, st. dev, 25th and 75th %iles) + NCE_monthly_mode: Lon, Lat, time and global 180x360x120 (LatxLonxMonth): NCE mode. + MAPS*: Lon, Lat and global 180x360 (LatxLon) datasets used in figures 1-3 in Bloom et al., (2015, in review). ##REFERENCES:## + Bloom AA, Williams M. (2015) Constraining ecosystem carbon dynamics in a data-limited world: integrating ecological' common sense' in a model-data fusion framework. Biogeosciences 12(5): 1299-1315. + Bloom et al., (2015, in review) The decadal state of the terrestrial carbon cycle: global constraints on terrestrial carbon allocation, pools and residence time. + Hiederer R, Kochy M (2011) Global Soil Organic Carbon Estimates and the Harmonized World Soil Database. EUR 25225 EN. Publications Office of the European Union. 79pp. + Saatchi SS, et al. (2011) Benchmark map of forest carbon stocks in tropical regions across three continents. Proc Natl Acad Sci 108(24): =9899-9904.

This upload contains datasets for benchmarking and improving different Sensor Fusion implementations/algorithms. The documentation for these datasets can be found on GitHub.

The upload contains two datasets (version 1.0.0):

• urban_with_gnss_dead_zones (7.0 GB, ~16 minutes)
  • City streets
  • A building is passed through on two occasions which makes the GNSS location signal unavailable at times.
  • RTK Fix is acquired at times
• suburban_nature (10.6 GB, ~19 minutes)
  • The route begins on a suburban street but quickly turns into a nature trail. Lots of vegetation
  • The RTK solution is only Float or None most of the route.

Details on collecting the data:

• Software
  • The data was collected using this open-source recorder.
    • Can be easily replayed using SpectacularAI's SDK (sdk-examples/python/oak/vio_replay.py)
  • Each dataset contains a map of the travelled route in Otaniemi, Espoo, Finland.
  • Necessary files to implement SLAM are included in the dataset.
  • Use of NTRIP and the high precision GNSS antenna enables global positioning accuracy of only few centimeters.
• Hardware
  • OAK-D stereo depth + color camera (Luxonis)
  • C099-F9P GNSS module (u-blox)
  • ANN-MB-00 high precision GNSS antenna (u-blox)

Metabolic profiling is routinely performed on multiple analytical platforms to increase the coverage of detected metabolites, and it is often necessary to distribute biological and clinical samples from a study between instruments of the same type to share the workload between different laboratories. The ability to combine metabolomics data arising from different sources is therefore of great interest, particularly for large-scale or long-term studies, where samples must be analyzed in separate blocks. This is not a trivial task, however, due to differing data structures, temporal variability, and instrumental drift. In this study, we employed blood serum and plasma samples collected from 29 subjects diagnosed with small cell lung cancer and analyzed each sample on two liquid chromatography–mass spectrometry (LC-MS) platforms. We describe a method for mapping retention times and matching metabolite features between platforms and approaches for fusing data acquired from both instruments. Calibration transfer models were developed and shown to be successful at mapping the response of one LC-MS instrument to another (Procrustes dissimilarity = 0.04; Mantel correlation = 0.95), allowing us to merge the data from different samples analyzed on different instruments. Data fusion was assessed in a clinical context by comparing the correlation of each metabolite with subject survival time in both the original and fused data sets: a simple autoscaling procedure (Pearson’s R = 0.99) was found to improve upon a calibration transfer method based on partial least-squares regression (R = 0.94).

Dataset collected in an indoor industrial environment using a mobile unit (manually pushed trolley) that resembles an industrial vehicle equipped with several sensors, namely, Wi-Fi, wheel encoder (displacement), and Inertial Measurement Unit (IMU).

Sensors were connected to a Raspberry Pi (RPi 3B +), which collected the data from the sensors. Ground truth information was obtained with video camera pointed towards the floor, registering the times when the trolley passed by reference tags.

List of sensors:

4x Wi-Fi interfaces: Edimax EW7811-Un

2x IMUs: Adafruit BNO055

1x Absolute Encoder: US Digital A2 (attached to a wheel with a diameter of 125 mm)

This dataset includes:

1x Wi-Fi radio map that can be used for Wi-Fi fingerprinting.

6x Trajectories: including sensor data + ground truth.

APs Information: list of APs in the building, including their position and transmission channel.

Floor plan: image of the building's floor plan with obstacles and non-navigable areas.

Python package provided for:

parsing the dataset into a data structure (Pandas dataframes).

performing statistical analysis on the data (number of samples, time difference between consecutive samples, etc.).

computing Dead Reckoning trajectory from a provided initial position.

computing Wi-Fi fingerprinting position estimates.

determining positioning error in Dead Reckoning and Wi-Fi fingerprinting.

generating plots including the floor plan of the building, dead reckoning trajectories, and CDFs.

When using this dataset, please cite its data description paper:

Silva , I.; Pendão, C.; Torres-Sospedra, J.; Moreira, A. Industrial Environment Multi-Sensor Dataset for Vehicle Indoor Tracking with Wi-Fi, Inertial and Odometry Data. Data 2023, 8, 157. https://doi.org/10.3390/data8100157

This dataset contains data for hand gestures recognition recorded with 3 different sensors. sEMG: recorded via the Myo armband that is composed of 8 equally spaced non-invasive sEMG sensors that can be placed approximately around the middle of the forearm. The sampling frequency of Myo is 200 Hz. The output of the Myo is a.u DVS: Dynamic Video Sensor which is a very low power event based camera with 128x128 resolution DAVIS: Dynamic Video Sensor which is a very low power event based camera with 240x180 resolution that also acquires APS frames. The dataset contains recordings of 10 subjects. Each subject performed 3 sessions, where each of the 5 hand gesture was recorded 5 times, each lasting for 2s. Between the gestures a relaxing phase of 1s is present where the muscles could go to the rest position, removing any residual muscular activation. Note: We did not upload the raw data (*.aedat) for the DAVIS being those files very heavy. All the information for the sensor has been extracted and can be found in the two files *.npz and *.mat. In case the raw data was needed please contact enea.ceolini@ini.uzh.ch elisa@ini.uzh.ch ==== README ==== DATASET STRUCTURE: EMG and DVS recordings 10 subjects 3 sessions for each subject 5 gestures in each session ('pinky', 'elle', 'yo', 'index', 'thumb') Data name: subjectXX_sessionYY_ZZZ XX : [01, 02, 03, 04, 05, 06, 07, 08, 09, 10] YY : [01, 02, 03] ZZZ : [emg, ann, dvs, davis] Data format: emg: .npy ann: .npy dvs: .aedat,.npy davis: .mat,.npz DVS DVS recordings only contain DVS events - .aedat (raw data): can be imported in Matlab using (https://github.com/inivation/AedatTools/tree/master/Matlab) or in Python with function aedat2numpy in converter.py (https://github.com/Enny1991/hand_gestures_cc19/tree/master/jAER_utils) - .npy (exported data): numpy.ndarray (xpos, ypos, ts, pol), 2D numpy array containing data of all events, timestamps ts reset to the trigger event (synchronized with the myo), timestamps ts in seconds DAVIS DAVIS recordings contain DVS events and APS frames. - .mat (exported data): mat structure, name 'aedat', events are inside aedat.data.polarity (aedat.data.polarity.x,aedat.data.polarity.y,aedat.data.polarity.timeStamp,aedat.data.polarity.polarity), aps frames are inside aedat.data.frame.samples, timestamps are in aedat.data.frame.timeStampStart (start of frame collection) or aedat.data.frame.timeStampEnd (end of frame collection) - .npz (exported data): npz files: ['frames_time', 'dvs_events', 'frames'], 'dvs_events' is a numpy.ndarray (xpos, ypos, ts, pol), 2D numpy array containing data of all events, timestamps ts reset to the trigger event (synchronized with the myo), timestamps ts in seconds; 'frames' and 'frames_time' are aps data, 'frames' is a list of all the frames, reset at the triggered time, 'frames_time' is the time for each frame, we considered the start timeStamps for each frame.

To improve the spatial and temporal reliability of LC patterns in ecologically heterogeneous regions, we adopted a multi-source data fusion approach based on the RF classification algorithm implemented on the GEE platform. RF is widely recognized for its robustness and ability to handle noisy or imbalanced training data,making it well-suited for LC mapping in complex and diverse landscapes. In this study, 30-meter LC maps for 2000, 2010, and 2020 were mapped by integrating multi-temporal Landsat imagery (±2 years) with training samples derived from the eight LC datasets.CodeType1Cropland2Forest3Shrub/Grassland4Water Bodies5Impervious Surfaces6Wetlands7Bare Land

The dataset contains raw visual images, visualized tactile images along the X- and Z-axes and an Excel file that organize every sample and their correspondence in order. The tactile images are interpolated on the raw haptic signal to align with the visual images. Both the visual and tactile images have identical resolution of 620 X 410. The dataset consists of 687 records. Each record includes one visual image, two tactile images along the X and Z axes, and one defect segmentation image. Tactile image filenames ending with x and z denote X and Z components respectively.The samples in the dataset exhibit a wide range of colors and textures. Moreover, the dataset demonstrates the advantage of cross-modal data fusion. As a flexible material, leather may have defects on its surface and underside, which can be observed in the visual and tactile images, respectively. Combining visual and tactile images provides better information on the distribution of defects

Introduction: We here describe a new method for distinguishing authentic Bletilla striata from similar decoctions (namely, Gastrodia elata, Polygonatum odoratum, and Bletilla ochracea schltr).Methods: Preliminary identification and analysis of four types of decoction pieces were conducted following the Chinese Pharmacopoeia and local standards. Intelligent sensory data were then collected using an electronic nose, an electronic tongue, and an electronic eye, and chromatography data were obtained via high-performance liquid chromatography (HPLC). Partial least squares discriminant analysis (PLS-DA), support vector machines (SVM), and back propagation neural network (BP-NN) models were built using each set of single-source data for authenticity identification (binary classification of B. striata vs. other samples) and for species determination (multi-class sample identification). Features were extracted from all datasets using an unsupervised approach [principal component analysis (PCA)] and a supervised approach (PLS-DA). Mid-level data fusion was then used to combine features from the four datasets and the effects of feature extraction methods on model performance were compared.Results and Discussion: Gas chromatography–ion mobility spectrometry (GC-IMS) showed significant differences in the types and abundances of volatile organic compounds between the four sample types. In authenticity determination, the PLS-DA and SVM models based on fused latent variables (LVs) performed the best, with 100% accuracy in both the calibration and validation sets. In species identification, the PLS-DA model built with fused principal components (PCs) or fused LVs had the best performance, with 100% accuracy in the calibration set and just one misclassification in the validation set. In the PLS-DA and SVM authenticity identification models, fused LVs performed better than fused PCs. Model analysis was used to identify PCs that strongly contributed to accurate sample classification, and a PC factor loading matrix was used to assess the correlation between PCs and the original variables. This study serves as a reference for future efforts to accurately evaluate the quality of Chinese medicine decoction pieces, promoting medicinal formulation safety.

A computational procedure to search for selective ligands for structurally related protein targets was developed and verified for serotonergic 5-HT7/5-HT1A receptor ligands. Starting from a set of compounds with annotated activity at both targets (grouped into four classes according to their activity: selective toward each target, not-selective and not-selective but active) and with an additional set of decoys (prepared using DUD methodology), the SVM (Support Vector Machines) models were constructed using a selective subset as positive examples and four remaining classes as negative training examples. Based on these four component models, the consensus classifier was then constructed using a data fusion approach. The combination of two approaches of data representation (molecular fingerprints vs. structural interaction fingerprints), different training set sizes and selection of the best SVM component models for consensus model generation, were evaluated to determine the optimal settings for the developed algorithm. The results showed that consensus models with molecular fingerprints, a larger training set and the selection of component models based on MCC maximization provided the best predictive performance.

Objective: The study aimed to develop simplified diagnostic models for identifying girls with central precocious puberty (CPP), without the expensive and cumbersome gonadotropin-releasing hormone (GnRH) stimulation test, which is the gold standard for CPP diagnosis.

Materials and Methods: Female patients who had secondary sexual characteristics before 8 years old and had taken a GnRH analog (GnRHa) stimulation test at a medical center in Guangzhou, China were enrolled. Data from clinical visiting, laboratory tests and medical image examinations were collected. We first extracted features from unstructured data such as clinical reports and medical images. Then, models based on each single-source data or multi-source data were developed with Extreme Gradient Boosting (XGBoost) classifier to classify patients as CPP or non-CPP.

Results: The best performance achieved an AUC of 0.88 and Youden index of 0.64 in the model based on multi-source data. The performance of single-source models ba...

Read first

User Guide is available at https://www.researchgate.net/profile/Yu-Xia-28/publications

If you used this dataset, please cite:
**"Yu Xia, Wei He, Qi Huang, Guoying Yin, Wenbin Liu, Hongyan Zhang, CRformer: Multi-modal data fusion to reconstruct cloud-free optical imagery, International Journal of Applied Earth Observation and Geoinformation, Volume 128, 2024, 103793, ISSN 1569-8432, https://doi.org/10.1016/j.jag.2024.103793.

"Y. Xia, W. He, Q. Huang, H. Chen, H. Huang and H. Zhang, "SOSSF: Landsat-8 Image Synthesis on the Blending of Sentinel-1 and MODIS Data," in IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1-19, 2024, Art no. 5401619, doi: 10.1109/TGRS.2024.3352662.**

@article{XIA2024103793, title = {CRformer: Multi-modal data fusion to reconstruct cloud-free optical imagery}, journal = {International Journal of Applied Earth Observation and Geoinformation}, volume = {128}, pages = {103793}, year = {2024}, issn = {1569-8432}, doi = {https://doi.org/10.1016/j.jag.2024.103793}, url = {https://www.sciencedirect.com/science/article/pii/S156984322400147X}, author = {Yu Xia and Wei He and Qi Huang and Guoying Yin and Wenbin Liu and Hongyan Zhang}}

@ARTICLE{10401024, author={Xia, Yu and He, Wei and Huang, Qi and Chen, Hongyu and Huang, He and Zhang, Hongyan}, journal={IEEE Transactions on Geoscience and Remote Sensing}, title={SOSSF: Landsat-8 Image Synthesis on the Blending of Sentinel-1 and MODIS Data}, year={2024}, volume={62}, number={}, pages={1-19}, keywords={Remote sensing;Earth;Artificial satellites;European Space Agency;Satellite constellations;MODIS;Optical sensors;Attention-based network;spatial–spectral fusion;synthetic aperture radar (SAR)–optical data;worldwide dataset}, doi={10.1109/TGRS.2024.3352662}}

Note If you can not click button "Donwload", you can download from the following "Data Explorer". The data set is a worldwide, multi-modal, real cloud removal dataset. We now release all sets, and RGB images of training set.

If you have any question or need some help, please contact the author's email address: whuxiayu@whu.edu.cn.

The SMILE dataset is for SAR-optical spatial-spectral fusion (SOSSF), obtained in 2019, open-sourced in https://www.kaggle.com/datasets/yuxiawhu/smile. The SMILE-CR data is a multi-modal data dataset for cloud removal in Landsat-8, which is created in 2020 with real clouds. It also can be applied to SOSSF task.

**Basic information **
The proposed SMILE-CR dataset consists of 1,400 image pairs of cloudy and clear images from the Landsat-8 sensor. The cloudy images are synthesized from real clouds captured at different times, with varying cloud cover ratios, ranging from no cloud cover to full cloud cover. Some examples of the dataset are displayed in Fig.~\ref{fig:fig3}. To facilitate the effective use of the dataset, it is randomly split into training, validation and testing sets, with 1,000, 200 and 200 patches respectively. The SMILE-CR dataset is compared with the state-of-the-art cloud removal dataset in Table~\ref{tabel:table2}. It is apparent that the SMILE-CR dataset is globally distributed with multi-modal data and accurate cloud masks, surpassing the existing cloud removal datasets. While most multi-modal cloud removal datasets only have two sensors, the SMILE-CR dataset has three sensors, enabling more image fusion tasks, such as image super-resolution and heterogenous spatial-spectral fusion. Moreover, the SMILE-CR dataset is one of the few datasets that provides accurate cloud masks, which can be used for both cloud removal and detection tasks. All in all, the SMILE-CR is a worldwide and curated benchmark dataset, bridging the gap of multi-modal cloud removal datasets in Landsat-8 sensor.

The dataset includes time-series data collected by four different sensors, which measure two target gases, Hydrogen Sulfide and Methyl Mercaptan, in the presence of air. To obtain measurements, each gas was individually exposed to the multi-sensor setup while in the presence of air. The dataset is particularly useful for gas classification tasks, as deep learning and data fusion techniques can be applied to identify the target gases.

The dataset comprises time-series data collected by four sensors, which measure two target gases, Hydrogen Sulfide (H2S) and Methyl Mercaptan (CH3SH) in the presence of air. To obtain measurements, each gas was individually exposed to the multi-sensor setup, while maintaining room temperature. Table 2 in the description file presents the distribution of data samples for the target gases collected from the multi-sensor system against the true gas concentration in parts per million (ppm) at two different humidity levels. The dataset file is available in CSV format and contains 9 columns with a total of 654440x4 gas samples. The CSV file also includes additional information on temperature, humidity, and true concentrations of Hydrogen Sulfide and Methyl Mercaptan. Out of the 654440x4 samples, there are 151682x4 samples of Methyl Mercaptan, 126142x4 samples of Hydrogen Sulfide, and the remaining samples are normal air samples.

The dataset was originally published in DiVA and moved to SND in 2024.