Loading...
Loading...
Image classification, object detection, segmentation, face recognition, OCR, image generation, video understanding
16,277 datasets
A dataset hosted on Kaggle, likely for training and evaluating object detection models. The specific content, size, and origin are not detailed in the provided metadata. Its name suggests it is related to the YOLO (You Only Look Once) family of computer vision architectures.
This dataset supports research on how citizens' perceptual accuracy of political parties' ideological positions changes over time, particularly around election campaigns. It leverages data from the Comparative Study of Electoral Systems across 21 established democracies and the British Election Study Internet Panel. The analysis examines fluctuations in accuracy linked to the electoral cycle.
The dataset appears to contain documents related to a legal case involving the Virginia Democratic Party and the U.S. Supreme Court. It was published on Kaggle. The specific content, time range, and author are unknown from the provided metadata.
EfficientNet-B0 weights published on Kaggle. The dataset likely contains the pre-trained parameters for the EfficientNet-B0 convolutional neural network architecture. Columns suggest it includes the weight matrices and biases necessary for image classification tasks.
ECGID_V2 suggests a version of an electrocardiogram (ECG) signal dataset. The title indicates the data likely contains noisy segments and references methods like PGD and GAN, which are used in adversarial machine learning and signal generation. Published on Kaggle, its specific content and scale require verification after download.
This panel dataset, authored by Tanay Bhatt and hosted on Harvard Dataverse, evaluates the impact of organized guarding on human mortality rates resulting from human-elephant conflict in Northeast India. It includes the final processed data files and R scripts used to generate regression tables and figures for the associated research. The data was last updated in March 2026.
An image dataset for training object detection models to identify damage on 13 distinct car parts. The raw description claims a model trained on this data can achieve 98% accuracy with a YOLOv8s or larger architecture. The dataset's author, organization, and specific collection details are unknown.
Surface water samples from the UK and Faroe Islands were analyzed for chemistry and organic matter composition between 2018 and 2020. The dataset, managed by the Environmental Information Data Centre, includes measurements for absorbance, elemental content, and other water chemistry parameters from streams, peat pools, and headwaters.
2020 data contains peat properties including moisture, bulk density, ash, and organic matter content from short cores collected 10 months post-fire. Samples were taken from high, medium, and low severity burn areas within drained and near-natural zones of a wildfire impacting over 6500 hectares of blanket bog and wet heath in Northern Scotland's Flow Country.
UK river water samples provide radiocarbon (14C) content and d13C measurements, along with suspended particulate matter concentration and organic carbon percentage. Data is concentrated from four major UK catchments—the Ribble, Conwy, Hampshire Avon, and Scottish Dee—and was collected during high flow conditions in 2013-2014.
Amazon Basin soil samples from 49 old-growth forest permanent plots provide measurements of pyrogenic carbon, organic carbon, and the ratio of %PyC to %Bulk Carbon. The dataset comprises 395 soil samples collected between 2015 and 2019. Measurements focus on soil fertility gradients within intact forests.
127,000 images form a modified subsample of the HAnd Gesture Recognition Image Dataset (HaGRID). The dataset is intended for classification tasks related to hand gestures. It is hosted on Kaggle and is associated with computer vision and human-computer interaction applications.
LibreYOLO provides an object detection dataset containing 717 annotated images across 28 classes of agricultural pests. The dataset is part of the Roboflow 100 benchmark, which spans 7 imagery domains. It was last updated on the Hugging Face platform in January 2026.
Real-time finger count detection for gesture control in elder care robotics. The dataset likely contains images or video frames annotated for hand detection and finger counting from zero to five. It is hosted on Kaggle and associated with the YOLOv8 object detection model.
ECGID V1 is a dataset of electrocardiogram (ECG) segments posted on Kaggle. The title suggests the data contains noise and was not augmented using PGD (Projected Gradient Descent) or GAN (Generative Adversarial Network) techniques. The dataset's author, organization, and specific scale are unknown.
A dataset focused on images for detecting outputs from Deep Convolutional Generative Adversarial Networks. It is hosted on Kaggle, but the specific creation date, author, and data volume are unknown. The dataset's primary purpose appears to be related to distinguishing between real and GAN-generated images.
ResNet50_Result is a dataset containing the output of a ResNet50 model, likely from Kaggle. The specific content, size, and creator are unknown. The dataset's last update date is also unknown.
EfficientNetB0_Result is a dataset published on Kaggle. It likely contains evaluation results from the EfficientNetB0 computer vision model. The specific content, such as performance metrics or predictions, must be verified after download.
Northern Michigan Chain of Lakes Dataset likely contains geospatial information about interconnected lakes in the region. Published on Kaggle, its specific contents, such as water quality measurements, depth soundings, or shoreline data, require verification after download. The dataset's author, organization, and last update date are unknown.
A collection of 2,101 aerial images containing 15,511 annotated elephants, split into training and test sets. The dataset was created by J. L. P. Naudé from images acquired over 8 separate campaigns in different environments. Image resolution varies between 2.4 cm/pixel and 13 cm/pixel, with nominal resolution specified in metadata.