Find research datasets worth reusing

opencc-by-4.0Feb 2019View details →

YCB-M: A Multi-Camera RGB-D Dataset for Object Recognition and 6DoF Pose Estimation

While a great variety of 3D cameras have been introduced in recent years, most publicly available datasets for object recognition and pose estimation focus on one single camera.  This dataset consists of 32 scenes that have been captured by 7 different 3D cameras, totaling 49,294 frames. This allows evaluating the sensitivity of pose estimation algorithms to the specifics of the used camera and the development of more robust algorithms that are more independent of the camera model. Vice versa, our dataset enables researchers to perform a quantitative comparison of the data from several different cameras and depth sensing technologies and evaluate their algorithms before selecting a camera for their specific task. The scenes in our dataset contain 20 different objects from the common benchmark YCB object and model set. We provide full ground truth 6DoF poses for each object, per-pixel segmentation, 2D and 3D bounding boxes and a measure of the amount of occlusion of each object. If you use this dataset in your research, please cite the following publication: T. Grenzdörffer, M. Günther, and J. Hertzberg, “YCB-M: A Multi-Camera RGB-D Dataset for Object Recognition and 6DoF Pose Estimation,” in 2020 IEEE International Conference on Robotics and Automation, ICRA 2020, Paris, France, May 31-June 4, 2020. IEEE, 2020. <pre><code>@InProceedings{Grenzdoerffer2020ycbm, title = {{YCB-M}: A Multi-Camera {RGB-D} Dataset for Object Recognition and {6DoF} Pose Estimation}, author = {Grenzd{\"{o}}rffer, Till and G{\"{u}}nther, Martin and Hertzberg, Joachim}, booktitle = {2020 {IEEE} International Conference on Robotics and Automation, {ICRA} 2020, Paris, France, May 31-June 4, 2020}, year = {2020}, publisher = {{IEEE}} }</code></pre> This paper is also available on arXiv: <a href="https://arxiv.org/abs/2004.11657">https://arxiv.org/abs/2004.11657</a>   To visualize the dataset, follow these instructions (tested on Ubuntu Xenial 16.04): <pre><code class="language-bash"># IMPORTANT: the ROS setup.bash must NOT be sourced, otherwise the following error occurs: # ImportError: /opt/ros/kinetic/lib/python2.7/dist-packages/cv2.so: undefined symbol: PyCObject_Type # nvdu requires Python 3.5 or 3.6 sudo add-apt-repository -y ppa:deadsnakes/ppa # to get python3.6 on Ubuntu Xenial sudo apt-get update sudo apt-get install -y python3.6 libsm6 libxext6 libxrender1 python-virtualenv python-pip # create a new virtual environment virtualenv -p python3.6 venv_nvdu cd venv_nvdu/ source bin/activate # clone our fork of NVIDIA's Dataset Utilities that incorporates some essential fixes pip install -e 'git+https://github.com/mintar/Dataset_Utilities.git#egg=nvdu' # download and transform the meshes # (alternatively, unzip the meshes contained in the dataset # to <path to venv_nvdu>/lib/python3.6/site-packages/nvdu/data/ycb/aligned_cm) nvdu_ycb -s # run nvdu_viz to visualize the dataset cd <a subdirectory of the YCB-M dataset with some frames> nvdu_viz --name_filters '*.jpg' </code></pre> For further details, see README.md.

opencc-by-4.0Aug 2024View details →

Visualization of the numerical pose optimization with the JAMDA scoring function using the BFGS and the LSL-BFGS algorithm

These videos demonstrate the behavior of two different optimization algorithms (BFGS and LSL-BFGS) during pose optimization using the JAMDA protein-ligand scoring function. Flachsenberg et al. (2020) (<a href="http://doi.org/10.1021/acs.jcim.0c01095" target="_blank" rel="noopener">10.1021/acs.jcim.0c01095</a>) describes the JAMDA protein-ligand scoring function and the LSL-BFGS algorithm. The data for these videos stems from Experiment 5 in Flachsenberg et al. (2020). In this experiment, the crystal structure of a ligand was numerically optimized in the binding site with respect to the JAMDA scoring function to create the JAMDA-minimized crystal structure. The JAMDA-minimized crystal structure was randomly deflected to generate various starting poses for the numerical optimization. These videos demonstrate the behavior of two optimization algorithms (BFGS and LSL-BFGS) when optimizing one of the generated starting poses. The chosen example for the videos is a structure of ribonuclease A with a 5'-deoxy-5'-N-piperidinouridine inhibitor (PDB code 3d6q, <a href="https://doi.org/10.2210/pdb3D6Q/pdb" target="_blank" rel="noopener">10.2210/pdb3D6Q/pdb</a>, <a href="https://doi.org/10.1021/jm800724t" target="_blank" rel="noopener">10.1021/jm800724t</a>). Each of the videos shows all the intermediate steps the optimization algorithm takes until convergence. The main observation (that is discussed in detail in Flachsenberg et al. (2020)) is that the BFGS algorithm tends to take inappropriately large steps when clashes are present in the structure, resulting in unwanted binding mode changes. This is not the case for the LSL-BFGS algorithm. <h3>Legend</h3> For each iteration, the JAMDA score value, the RMSD to the JAMDA-minimized crystal structure (yellow), and the RMSD to the optimization's starting point (blue) are given. Furthermore, the gradient's norm (representing the main convergence criterion) is shown. In addition to the optimized ligand, also the JAMDA-minimized crystal structure (yellow) and the optimization's starting structure (blue) are shown. Each optimization algorithm is shown in two videos: In one video, the optimized ligand is colored by elements. Here, atoms with clashes (positive JAMDA scores) are marked with orange balls. In the other video, the atoms and bonds of the optimized ligand are colored by their individual JAMDA score. <h3>Used Software</h3> The snapshots of the optimization algorithms were rendered using PyMOL 3.0 (<a href="https://pymol.org/" target="_blank" rel="noopener">https://pymol.org/</a>) and further processed using the Pillow 10.4 Python library (<a href="https://doi.org/10.5281/zenodo.12606429" target="_blank" rel="noopener">10.5281/zenodo.12606429</a>). Videos were created from the individual snapshots using FFmpeg 7.0 (<a href="https://www.ffmpeg.org/" target="_blank" rel="noopener">https://www.ffmpeg.org/</a>).

dryad40/100

Data from: Ancestral hybridization yields evolutionary distinct hybrids lineages and species boundaries in crocodiles, posing unique conservation conundrums

Interspecific hybridization can lead to adaptation and speciation, especially in the context of recent radiations. The emblematic Crocodylus (true crocodiles) is the most broadly distributed, ecologically diverse, and species-rich crocodylian genus. Nonetheless, their within-species evolutionary processes are poorly resolved mainly due to their potential for hybridization. Notably, the evolutionary outcomes when hybridization is ancient and involves long-lived species, like crocodiles, remain largely unexplored. Here, we evaluate the genomic admixture between the American (Crocodylus acutus) and the Morelet's (Crocodylus moreletii) species, and demonstrate that this hybridization system challenges the definition of species boundaries and poses a triple conservation conundrum: what has been recognized as C. acutus is actually two distinct species, therefore its taxonomic reassessment is needed; we identified two evolutionary distinct hybrids lineages, which are genetically discernible from the parental species; the remaining C. moreletii populations evidence its likely extinction as a species and/or evolution via hybridization. Hence, the crocodiles' distinct species and hybrids lineages warrant recognition and need urgent conservation efforts.

opencc-zeroDec 2017View details →

opencc-by-4.0Dec 2022View details →

Poses of People in Art: A Data Set for Human Pose Estimation in Digital Art History

Throughout the history of art, the pose—as the holistic abstraction of the human body's expression—has proven to be a constant in numerous studies. However, due to the enormous amount of data that so far had to be processed by hand, its crucial role to the formulaic recapitulation of art-historical motifs since antiquity could only be highlighted selectively. This is true even for the now automated estimation of human poses, as domain-specific, sufficiently large data sets required for training computational models are either not publicly available or not indexed at a fine enough granularity. With the Poses of People in Art data set, we introduce the first openly licensed data set for estimating human poses in art and validating human pose estimators. It consists of 2,454 images from 22 art-historical depiction styles, including those that have increasingly turned away from lifelike representations of the body since the 19th century. A total of 10,749 human figures are precisely enclosed by rectangular bounding boxes, with a maximum of four per image labeled by up to 17 keypoints; among these are mainly joints such as elbows and knees. For machine learning purposes, the data set is divided into three subsets—training, validation, and testing—, that follow the established JSON-based Microsoft COCO format, respectively. Each image annotation, in addition to mandatory fields, provides metadata from the art-historical online encyclopedia WikiArt.

opencc-by-4.0Mar 2023View details →

An Occlusion and Pose Sensitive Image Dataset for Black Ear Recognition

RESEARCH APPROACH The research approach adopted for the study consists of seven phases which includes as shown in Figure 1: <ol> <li>Pre-acquisition</li> <li>data pre-processing</li> <li>Raw images collection</li> <li>Image pre-processing</li> <li>Naming of images</li> <li>Dataset Repository</li> <li>Performance Evaluation</li> </ol> The different phases in the study are discussed in the sections below.   PRE-ACQUISITION The volunteers are given brief orientation on how their data will be managed and used for research purposes only. After the volunteers agrees, a consent form is given to be read and signed. The sample of the consent form filled by the volunteers is shown in Figure 1. The capturing of images was started with the setup of the imaging device. The camera is set up on a tripod stand in stationary position at the height 90 from the floor and distance 20cm from the subject.   EAR IMAGE ACQUISITION Image acquisition is an action of retrieving image from an external source for further processing. The image acquisition is purely a hardware dependent process by capturing unprocessed images of the volunteers using a professional camera. This was acquired through a subject posing in front of the camera. It is also a process through which digital representation of a scene can be obtained. This representation is known as an image and its elements are called pixels (picture elements). The imaging sensor/camera used in this study is a Canon E0S 60D professional camera which is placed at a distance of 3 feet form the subject and 20m from the ground.  This is the first step in this project to achieve the project’s aim of developing an occlusion and pose sensitive image dataset for black ear recognition. (OPIB ear dataset). To achieve the objectives of this study, a set of black ear images were collected mostly from undergraduate students at a public University in Nigeria.   The image dataset required is captured in two scenarios: 1. uncontrolled environment with a surveillance camera <ol> </ol> The image dataset captured is purely black ear with partial occlusion in a constrained and unconstrained environment.   2. controlled environment with professional cameras The ear images captured were from black subjects in controlled environment. To make the OPIB dataset pose invariant, the volunteers stand on a marked positions on the floor indicating the angles at which the imaging sensor was captured the volunteers’ ear. The capturing of the images in this category requires that the subject stand and rotates in the following angles 60o, 30o and 0o towards their right side to capture the left ear and then towards the left to capture the right ear (Fernando et al., 2017) as shown in Figure 4. Six (6) images were captured per subject at angles 60o, 30o and 0o for the left and right ears of 152 volunteers making a total of 907 images (five volunteers had 5 images instead of 6, hence folders 34, 22, 51, 99 and 102 contain 5 images). To make the OPIB dataset occlusion and pose sensitive, partial occlusion of the subject’s ears were simulated using rings, hearing aid, scarf, earphone/ear pods, etc. before the images are captured.   <table> <tbody> <tr> <td> CONSENT FORM This form was designed to obtain participant’s consent on the project titled: An Occlusion and Pose Sensitive Image Dataset for Black Ear Recognition (OPIB). The information is purely needed for academic research purposes and the ear images collected will curated anonymously and the identity of the volunteers will not be shared with anyone. The images will be uploaded on online repository to aid research in ear biometrics. The participation is voluntary, and the participant can withdraw from the project any time before the final dataset is curated and warehoused. Kindly sign the form to signify your consent. I consent to my image being recorded in form of still images or video surveillance as part of the OPIB ear images project. Tick as appropriate: GENDER       Male    Female AGE               (18-25)                        (26-35)                        (36-50)   ……………………………….. SIGNED </td> </tr> </tbody> </table>   Figure 1: Sample of Subject’s Consent Form for the OPIB ear dataset   RAW IMAGE COLLECTION The ear images were captured using a digital camera which was set to JPEG because if the camera format is set to raw, no processing will be applied, hence the stored file will contain more tonal and colour data. However, if set to JPEG, the image data will be processed, compressed and stored in the appropriate folders.   IMAGE PRE-PROCESSING The aim of pre-processing is to improve the quality of the images with regards to contrast, brightness and other metrics. It also includes operations such as: cropping, resizing, rescaling, etc. which are important aspect of image analysis aimed at dimensionality reduction. The images are downloaded on a laptop for processing using MATLAB.   Image Cropping The first step in image pre-processing is image cropping. Some irrelevant parts of the image can be removed, and the image Region of Interest (ROI) is focused. This tool provides a user with the size information of the cropped image. MATLAB function for image cropping realizes this operation interactively by waiting for a user to specify the crop rectangle with the mouse and operate on the current axes. The output images of the cropping process are of the same class as the input image. Naming of OPIB Ear Images The OPIB ear images were labelled based on the naming convention formulated from this study as shown in Figure 5. The images are given unique names that specifies the subject, the side of the ear (left or right) and the angle of capture. The first and second letters (SU) in the image names is block letter simply representing subject for subject 1-to-n in the dataset, while the left and right ears is distinguished using L1, L2, L3 and R1, R2, R3 for angles 600, 300 and 00, respectively as shown in Table 1.   Table 1: Naming Convention for OPIB ear images <table align="center"> <tbody> <tr> <td> NAMING CONVENTION </td> </tr> <tr> <td> Label Degrees                                     600     300      00 </td> </tr> <tr> <td> No of the degree                     1          2          3 </td> </tr> <tr> <td> Subject 1   indicates          (first image in dataset) SU1 </td> </tr> <tr> <td> Subject n   indicates          (last image in dataset) SUn </td> </tr> <tr> <td> Left Image 1                                 L 1 Left image n                                 L n Right Image 1                               R 1 Right Image n                               R n </td> </tr> <tr> <td> SU1L1                                           SU1RI SU1L2                                           SU1R2             SU1L3                                           SU1R3 </td> </tr> </tbody> </table>   OPIB EAR DATASET EVALUATION The prominent challenges with the current evaluation practices in the field of ear biometrics are the use of different databases, different evaluation matrices, different classifiers that mask the feature extraction performance and the time spent developing framework (Abaza et al., 2013; Emer&scaron;ič et al., 2017). The toolbox provides environment in which the evaluation of methods for person recognition based on ear biometric data is simplified. It executes all the dataset reads and classification based on ear descriptors.   DESCRIPTION OF OPIB EAR DATASET OPIB ear dataset was organised into a structure with each folder containing 6 images of the same person. The images were captured with both left and right ear at angle 0, 30 and 60 degrees. The images were occluded with earing, scarves and headphone etc.  The collection of the dataset was done both indoor and outdoor.  The dataset was gathered through the student at a public university in Nigeria. The percentage of female (40.35%) while Male (59.65%).  The ear dataset was captured through a profession camera Nikon D 350. It was set-up with a camera stand where an individual captured in a process order. A total number of 907 images was gathered. The challenges encountered in term of gathering students for capturing, processing of the images and annotations. The volunteers were given a brief orientation on what their ear could be used for before, it was captured, for processing.  It was a great task in arranging the ear (dataset) into folders and naming accordingly.   Table 2: Overview of the OPIB Ear Dataset <table align="left"> <tbody> <tr> <td> Location </td> <td> Both Indoor and outdoor environment </td> </tr> <tr> <td> Information about Volunteers </td> <td> Students </td> </tr> <tr> <td> Gender </td> <td> Female (40.35%) and male (59.65%) </td> </tr> <tr> <td> Head Side Left and Right </td> <td> Side Left and Right </td> </tr> <tr> <td> Total number of volunteers </td> <td> 152 </td> </tr> <tr> <td> Per Subject images </td> <td> 3 images of left ear and 3 images of right ear </td> </tr> <tr> <td> Total Images </td> <td> 907 </td> </tr> <tr> <td> Age group </td> <td> 18 to 35 years </td> </tr> <tr> <td> Colour Representation </td> <td> RGB </td> </tr> <tr> <td> Image Resolution </td> <td> 224x224 </td> </tr> </tbody> </table>

opencc-by-4.0Nov 2022View details →

HoloSet - A Dataset for Visual-Inertial Pose Estimation in Extended Reality

HoloSet was published at DATA'22 workshop at ACM SenSys'22.  Overview: There is a lack of datasets for visual-inertial odometry applications in Mixed Reality (MR). To the best of our knowledge, there is no dataset available that is captured from an MR headset with a human as a carrier. To bridge this gap, we present a novel pose estimation dataset — called HoloSet — collected using Microsoft Hololens 2, which is a state-of-the-art head mounted device for XR. Potential applications for HoloSet include visual-inertial odometry, simultaneous localization and mapping (SLAM), and additional applications in XR that leverage visual-inertial data. HoloSet captures both macro and micro movements. For macro movements, the dataset consists of more than 66,000 samples of visual, inertial, and depth camera data in a variety of environments (indoor, outdoor) and scene setups (trails, suburbs, downtown) under multiple user action scenarios (walk, jog). For micro movements, the dataset consists of more than 12,000 samples of additional articulated hand depth camera images while a user plays games that exercise fine motor skills and hand-eye coordination. We present basic visualizations and high-level statistics of the data and outline the potential research use cases for HoloSet. Please find the relevant publication at https://dl.acm.org/doi/abs/10.1145/3560905.3567763.

dryad40/100

Validating marker-less pose estimation with 3D x-ray radiography

Open the record for dataset details and reuse information.

publicMay 2022View details →

dryad40/100

Data from: Ancestral hybridization yields evolutionary distinct hybrids lineages and species boundaries in crocodiles, posing unique conservation conundrums

Open the record for dataset details and reuse information.

publicAug 2019View details →

opencc-zeroAug 2020View details →

Agricultural intensification heightens food safety risks posed by wild birds

1. Agricultural intensification and simplification are key drivers of recent declines in wild bird populations, heightening the need to better balance conservation with food production. This is hindered, however, by perceptions that birds threaten food safety. While birds are known reservoirs of foodborne pathogens, there remains uncertainty about the links between landscape context, farming practices, and actual crop contamination by birds. 2. Here, we examine relationships between landscape context, farming practices, and pathogen contamination by birds using a barrier-to-spillover approach. First, we censused bird communities using point count surveys. Second, we collected 2024 faecal samples from captured birds alongside 1215 faecal samples from brassica fields and food processing areas across 50 farms spanning the USA West Coast. We then estimated the prevalence of three foodborne pathogens across landscape and livestock intensification gradients. Finally, we quantified the number of plants with faeces. 3. Campylobacter spp. were detected in 10.2% of faeces from captured birds and 13.1% of faeces from production areas. Nonnative birds were 4.1 times more likely to have Campylobacter spp. than native birds. Salmonella spp. were detected in 0.2% of faeces from production areas and were never detected in captured birds. We detected evidence of Shiga toxigenic E. coli in 1 sample across the >3200 tested. 4. Campylobacter spp. prevalence in faeces from production areas increased with increasing mammalian livestock densities in the landscape but decreased with increasing amounts of natural habitat. 5. We encountered bird faeces on 3.3% of plants examined. Despite the impact on pathogen prevalence, landscape context did not increase the number of plants with bird faeces, although on-farm mammalian livestock density slightly did. 6. Synthesis and applications. Food safety and wildlife conservation are often thought to be in conflict. However, our findings suggest that natural habitat around farms may reduce crop contamination rates by birds. This is perhaps because natural habitat can promote native birds that are less likely to harbour foodborne pathogens or because it decreases contact with livestock waste. Our results suggest that preservation of natural habitats around farms could benefit both conservation and food safety, contrary to current standards for "best practices."

opencc-zeroDec 2017View details →

Data from: Deforestation risks posed by oil palm expansion in the Peruvian Amazon

Further expansion of agriculture in the tropics is likely to accelerate the loss of biodiversity. One crop of concern to conservation is African oil palm (Elaeis guineensis). We examined recent deforestation associated with oil palm in the Peruvian Amazon within the context of the region's other crops. We found more area under oil palm cultivation (845 km2 ) than did previous studies. While this comprises less than 4% of the cropland in the region, it accounted for 11% of the deforestation from agricultural expansion from 2007 to 2013. Patches of oil palm agriculture were larger and more spatially clustered than for other crops, potentially increasing their impact on local habitat fragmentation. Modeling deforestation risk for oil palm expansion using climatic and edaphic factors showed that sites at lower elevations, with higher precipitation, and lower slopes than those typically used for intensive agriculture are at long-term risk of deforestation from oil palm agriculture. Within areas at long-term risks, based on CART models, areas near urban centers, roads, and previously deforested areas are at greatest short-term risk of deforestation. Existing protected areas and officially recognized indigenous territories cover large areas at long-term risk of deforestation for oil palm (>40%). Less than 7% of these areas are under strict (IUCN I-IV) protection. Based on these findings, we suggest targeted monitoring for oil palm deforestation as well as strengthening and expanding protected areas to conserve specific habitats.

opencc-zeroOct 2023View details →

mRI: multi-modal 3d human pose estimation dataset using mmwave, rgb-d, and inertial sensors

The ability to estimate 3D human body pose and movement, also known as human pose estimation~(HPE), enables many applications for home-based health monitoring, such as remote rehabilitation training. Several possible solutions have emerged using sensors ranging from RGB cameras, depth sensors, millimeter-Wave (mmWave) radars, and wearable inertial sensors. Despite previous efforts on datasets and benchmarks for HPE, few datasets exploit multiple modalities and focus on home-based health monitoring. To bridge this gap, we present mRI, a multi-modal 3D human pose estimation dataset with mmWave, RGB-D, and Inertial Sensors. Our dataset consists of over 5 million frames from 20 subjects performing rehabilitation exercises and supports the benchmarks of HPE and action detection. We perform extensive experiments using our dataset and delineate the strength of each modality. We hope that the release of mRI can catalyze the research in pose estimation, multi-modal learning, and action understanding, and more importantly, facilitate the applications of home-based health monitoring.

opencc-by-4.0Oct 2023View details →

Multi-modal pose estimation in XR applications leveraging integrated sensing and communication: Dataset

This dataset refers to paper Multi-modal pose estimation in XR applications leveraging integrated sensing and communication in workshop of ACM Mobicom. CSI of 3 people performing a set of 8 poses. This dataset contain discrete classes and corresponding CSI data. Kinect poses can be found here (https://github.com/nisarnabeel/multi-modal-pose-estimation-CSI-mmWave).   Abstract: Mobile extended reality (XR) applications are anticipated to generate substantial traffic for 6G. Such applications not only require high data rate and low-latency transmissions, but also accurate and real-time pose estimation to enable interactive and immersive experiences. While sub-6 GHz signals have been exploited for pose estimation, they cannot cope up with multi-gigabit data rates required by XR applications. Instead, mobile communications at mmWave frequencies can potentially support data rates up to several giga-bits per second (Gbps) and, therefore, can be used to deliver XR content wirelessly to the Head-Mounted Display (HMD). Moreover, mmWave frequencies can offer improved sensing due to the large available bandwidth. Therefore, mmWave communications can play a crucial role in enabling device-free interactivity by offering both high-speed communication and accurate sensing capabilities. However, mmWave propagation characteristics are different from sub-6 GHz. Path loss plays a significant role, and can lead to degraded sensing performance. Therefore, our proposal supplements wireless sensing at mmWave frequencies with wireless electromyography (EMG) armbands. By capturing patterns of muscle activities, we can counteract the limitations of mmWave-based pose estimation, thereby enriching the granularity and precision of pose estimation. This paper proposes a conceptual architecture to achieve multi-modal pose estimation for XR applications. Early results highlight the shortcomings of mmWave-based sensing, and we identify future steps and opportunities on integration of both approaches.

opencc-by-4.0Nov 2023View details →

6DAPose - Synthetic Assembly Pose Dataset

6DAPose dataset contains an object assembly in an ordered manner following its  assembly steps.  Each dataset contains 431 simulated view samples of the assembly for each assembly step. For each view sample, RGB and Depth Image, Segmentation map, ground truth values for 6D object pose and camera pose and model information are recorded following the <a href="https://bop.felk.cvut.cz/home/">BOP format </a>. There are two datasets representing a 3-step fidget gear assembly and a 4-step Nema17 gear reducer assembly.

opencc-byMay 2022View details →

Neanderthal Composite - A-Pose

Neanderthals were ancient humans who lived in Eurasia. Many specifically Neanderthal traits could be explained by adaptation to cold environments. XR viewing capable. Model made by Keith Chan using Blender. The model includes teeth and a Mousterian point hafted on a spear. The whole dimensions are 1.19m x 0.324m x 1.65m. See this model and more in life size at [AnVRopomotron.com](http://www.anvropomotron.com). Source: Objaverse 1.0 / Sketchfab

opencc-by-4.0Feb 2024View details →

CSI4Free: GAN-Augmented mmWave CSI for Improved Pose Classification

This refers to the gan-generated dataset for the paper" CSI4Free: GAN-Augmented mmWave CSI for Improved Pose Classification. Abstract:In recent years, Joint Communication and Sensing (JC&S), has demonstrated significant success, particularly in utilizing sub-6 GHz frequencies with commercial-off-the-shelf (COTS) Wi-Fi devices for applications such as localization, gesture recognition, and pose classification. Deep learning and the existence of large public datasets has been pivotal in achieving such results. However, at mmWave frequencies (30-300 GHz), which has shown potential for more accurate sensing performance, there is a noticeable lack of research in the domain of COTS Wi-Fi sensing. Challenges such as limited research hardware, the absence of large datasets, limited functionality in COTS hardware, and the complexities of data collection present obstacles to a comprehensive exploration of this field. In this work, we aim to address these challenges by developing a method that can generate synthetic mmWave channel state information (CSI) samples. In particular, we use a generative adversarial network (GAN) on an existing dataset, to generate 30,000 additional CSI samples. The augmented samples exhibit a remarkable degree of consistency with the original data, as indicated by the notably high GAN-train and GAN-test scores. Furthermore, we integrate the augmented samples in training a pose classification model. We observe that the augmented samples complement the real data and improve the generalization of the classification model. The repository is available here: https://github.com/nisarnabeel/Dataset-GAN-Augmented-mmWave-CSI-for-improved-pose-classification   Paper Link:https://ieeexplore.ieee.org/document/10646223/

opencc-by-4.0Mar 2024View details →

Problem Posing Booklet

Open the record for dataset details and reuse information.

opencc-by-4.0Apr 2024View details →

INTELLIMAN_WP2_Application Requirements and Integration_T2.4_Fresh food handling use case analysis, integration and validation_Apple 6D pose estimation dataset_v0

The dataset contains the data generated for the training of the 6D pose estimation neural network DOPE related to the publication: M. Costanzo, M. De Simone, S. Federico, C. Natale and S. Pirozzi, "Enhanced 6D Pose Estimation for Robotic Fruit Picking," 2023 9th International Conference on Control, Decision and Information Technologies (CoDIT), Rome, Italy, 2023, pp. 901-906, doi: 10.1109/CoDIT58514.2023.10284072.

opencc-by-4.0Nov 2021View details →

MODYS-video: 2D Human pose estimation data and Dyskinesia Impairment Scale scores from children and young adults with dyskinetic cerebral palsy

The dataset contains the 2D coordinates in pixels of body landmarks (wrists, ankles, shoulders, hips, knees and ankles) extracted from 188 videos of 34 children with dyskinetic cerebral palsy using DeepLabCut [1] and appertaining clinical scores of the Dyskinesia Impairment Scale (DIS) [2]. The videos were collected during the item “lying in rest” and “sitting in rest” of the DIS at three time points during a clinical trial on the effect of intrathecal baclofen [3]. Children had a mean age of 14y2m (SD 4.0), 26 were male. Their gross motor function classification system level ranged from IV-V and their manual ability classification system level from III-V. Original videos have length of 4-35 seconds with a resolution of 720x575 pixels and are sampled with 25 Hz. We added stick figures to complement the data for context and ease of understanding. They were created from the 2D coordinates that were extracted with a likelihood >0.8. Clinical scoring was performed by three trained experts (according to the DIS) on the original videos. Within the items “lying in rest” and “sitting in rest” the amplitude and duration of dystonia and choreoathetosis of the trunk, proximal right arm, proximal left arm, proximal right leg and proximal left leg are scored on a 0-4 ordinal scale and calculated towards a percentage score between 0-1. The dataset can be used in a machine learning approach to automatically assess dystonia and choreoathetosis of children with dyskinetic cerebral palsy using 2D coordinates of body points extracted from videos.   References: 1. Mathis, A., et al., DeepLabCut: markerless pose estimation of user-defined body parts with deep learning. Nat Neurosci, 2018. 21(9): p. 1281-1289. 2. Monbaliu, E., et al., The dyskinesia Impairment Scale: a new instrument to measure dystonia and choreoathetosis in dyskinetic cerebral palsy. Dev Med Child Neurol, 2012. 54: p. 278-283. 3. Bonouvrie, L.A., et al., The Effect of Intrathecal Baclofen in Dyskinetic Cerebral Palsy: The IDYS Trial. Ann Neurol, 2019. 86: p. 79-90.

opencc-by-nc-4.0Apr 2022View details →

Tango Spacecraft Dataset for Monocular Pose Estimation

Reference Paper: <a href="https://doi.org/10.1016/j.actaastro.2023.01.012">M. Bechini, M. Lavagna, P. Lunghi, Dataset generation and validation for spacecraft pose estimation via monocular images processing, Acta Astronautica 204 (2023) 358–369</a> <a href="https://www.researchgate.net/publication/361924362_Spacecraft_Pose_Estimation_via_Monocular_Image_Processing_Dataset_Generation_and_Validation">M. Bechini, P. Lunghi, M. Lavagna. "Spacecraft Pose Estimation via Monocular Image Processing: Dataset Generation and Validation". In 9th European Conference for Aeronautics and Aerospace Sciences (EUCASS)</a> General Description: The "Tango Spacecraft Dataset for Monocular Pose Estimation" dataset here published should be used for relative pose estimation tasks. It is split into 30002 train images and 3002 test images representing the Tango spacecraft from Prisma mission, being the largest publicly available dataset of synthetic space-borne noise-free images tailored to pose estimation tasks (up to our knowledge). The label of each image gives relative quaternion (in scalar-last format) between Tango and the camera (hence the relative position of the target with respect to the camera in camera reference frame) and the relative position of Tango with respect to the camera in camera reference frame. More information on the dataset split and on the label format are reported below.  Images Information: The dataset comprises 30002 synthetic grayscale images of Tango spacecraft from Prisma mission that serves as train set, while the test set is formed by 3002 synthetic grayscale images of Tango spacecraft from Prisma mission in PNG format. About 1/6 of the images both in the train and in the test set have a non-black background, obtained by rendering an Earth-like model in the raytracing process used to define the images reported. The images are noise-free to increase the flexibility of the dataset. The illumination direction of the spacecraft in the scene is uniformly distributed in the 3D space in agreement with the Sun position constraints. The dataset contains also a .txt file with the parameters of the camera used to generate the images. Labels Information: Labels in the SPEED and SPEED+ dataset format are here provided in separated JSON files. The files are formatted per each image as in the following example: <ul> <li>    filename              : tango_img_1                               # name of the image to which the data are referred</li> <li>    q_TRG2CAM      : [qx qy qz qw]                               # relative quaternion from Target to Camera reference frame</li> <li>    t_CAM2TRG       : [x, y, z]                                         # relative position of Tango with respect to the camera expressed in meters</li> </ul> Notice that for making the usage of the dataset easier, both the training set and the test set are split in two folders containing the images with earth as background and without background. VERSION CONTROL <ul> <li>v1.0: This version contains the dataset (both train and test) of full scale images with relative pose annotations. These images have width=height=1024 pixels. The position of tango with respect to the camera is randomly selected from a uniform distribution, but it is ensured the full visibility in all the images. </li> </ul> Note: this dataset contains the same images of the "Tango Spacecraft Wireframe Dataset Model for Line Segments Detection" v2.0 full-scale (DOI: <a href="https://doi.org/10.5281/zenodo.6372848">https://doi.org/10.5281/zenodo.6372848</a>) and also "Tango Spacecraft Dataset for Region of Interest Estimation and Semantic Segmentation" v1.0 (DOI: <a href="https://doi.org/10.5281/zenodo.6507863">https://doi.org/10.5281/zenodo.6507863</a>) and they can be used together by combining the annotations of the relative pose and the ones of the reprojected wireframe model of Tango, with also the ones of the ROI. These three datasets give the most comprehensive dataset of space borne synthetic images ever published (up to our knowledge).