nvBench is a large-scale NL2VIS (natural languagge to visualisations) benchmark, containing 25,750 (NL, VIS) pairs from 750 tables over 105 domains, synthesized from (NL, SQL) benchmarks to support cross-domain NLPVIS (Natural Language Query to Visualization) task.
8 PAPERS • NO BENCHMARKS YET
We introduce ArtBench-10, the first class-balanced, high-quality, cleanly annotated, and standardized dataset for benchmarking artwork generation. It comprises 60,000 images of artwork from 10 distinctive artistic styles, with 5,000 training images and 1,000 testing images per style. ArtBench-10 has several advantages over previous artwork datasets. Firstly, it is class-balanced while most previous artwork datasets suffer from the long tail class distributions. Secondly, the images are of high quality with clean annotations. Thirdly, ArtBench-10 is created with standardized data collection, annotation, filtering, and preprocessing procedures. We provide three versions of the dataset with different resolutions (32×32, 256×256, and original image size), formatted in a way that is easy to be incorporated by popular machine learning frameworks.
7 PAPERS • 1 BENCHMARK
DIOR-RSVG is a large-scale benchmark dataset of remote sensing data (RSVG). It aims to localize the referred objects in remote sensing (RS) images with the guidance of natural language. This new dataset includes image/expression/box triplets for training and evaluating visual grounding models.
7 PAPERS • NO BENCHMARKS YET
DUDE is formulated as an instance of Document Question Answering (DocQA) to evaluate how well current solutions deal with multi-page documents, if they can navigate and reason over the layout, and if they can generalize these skills to different document types and domains. Since we cannot provide question-answer pairs about, e.g., ticked checkboxes, on each document instance or document type, the challenge presented by DUDE is characterized equally as a Multi-Domain Long-Tailed Recognition problem
Although deep face recognition has achieved impressive results in recent years, there is increasing controversy regarding racial and gender bias of the models, questioning their trustworthiness and deployment into sensitive scenarios. DemogPairs is a validation set with 10.8K facial images and 58.3M identity verification pairs, distributed in demographically-balanced folds of Asian, Black and White females and males. We also propose a benchmark of experiments using DemogPairs over state-of-the-art deep face recognition models in order to analyze their cross-demographic behavior and potential demographic biases (see figure below).
A GQA-based dataset with 1,040,830 multi-modal explanations of visual reasoning processes.
A new large scale plane geometry problem solving dataset called PGPS9K, labeled both fine-grained diagram annotation and interpretable solution program.
ReaSCAN is a synthetic navigation task that requires models to reason about surroundings over syntactically difficult languages.
English subset of the SLAKE dataset, comprising 642 images and more than 7,000 question–answer pairs.
SpaceNet 2: Building Detection v2 - is a dataset for building footprint detection in geographically diverse settings from very high resolution satellite images. It contains over 302,701 building footprints, 3/8-band Worldview-3 satellite imagery at 0.3m pixel res., across 5 cities (Rio de Janeiro, Las Vegas, Paris, Shanghai, Khartoum), and covers areas that are both urban and suburban in nature. The dataset was split using 60%/20%/20% for train/test/validation.
Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset.
VISUELLE is a repository build upon the data of a real fast fashion company, Nunalie, and is composed of 5577 new products and about 45M sales related to fashion seasons from 2016-2019. Each product in VISUELLE is equipped with multimodal information: its image, textual metadata, sales after the first release date, and three related Google Trends describing category, color and fabric popularity.
The VNHSGE (VietNamese High School Graduation Examination) dataset, developed exclusively for evaluating large language models (LLMs), is introduced in this article. The dataset, which covers nine subjects, was generated from the Vietnamese National High School Graduation Examination and comparable tests. 300 literary essays have been included, and there are over 19,000 multiple-choice questions on a range of topics. The dataset assesses LLMs in multitasking situations such as question answering, text generation, reading comprehension, visual question answering, and more by including both textual data and accompanying images. Using ChatGPT and BingChat, we evaluated LLMs on the VNHSGE dataset and contrasted their performance with that of Vietnamese students to see how well they performed. The results show that ChatGPT and BingChat both perform at a human level in a number of areas, including literature, English, history, geography, and civics education. They still have space to grow, t
7 PAPERS • 9 BENCHMARKS
This dataset provides a new split of VQA v2 (similarly to VQA-CP v2), which is built of questions that are hard to answer for biased models.
ViQuAE is a dataset for KVQAE (Knowledge-based Visual Question Answering about named Entities), a task which consists in answering questions about named entities grounded in a visual context using a Knowledge Base. It is the first KVQAE dataset to cover a wide range of entity types (e.g. persons, landmarks, and products). We argue that KVQAE is a clear, well-defined task that can be evaluated easily, making it suitable to track the progress of multimodal entity representation’s quality. Multimodal entity representation is a central issue that will allow to make human-machine interactions more natural. For example, while watching a movie, one might wonder ‘‘Where did I already see this actress?’’ or ‘‘Did she ever win an Oscar?’’
Who's Waldo is a dataset of 270K image–caption pairs, depicting interactions of people, that is automatically mined from Wikimedia Commons. It is a benchmark dataset for person-centric visual grounding, the problem of linking between people named in a caption and people pictured in an image.
3DOH50K is the first real 3D human dataset for the problem of human reconstruction and pose estimation in occlusion scenarios. It contains 51600 images with accurate 2D pose and 3D pose, SMPL parameters, and binary mask.
6 PAPERS • 1 BENCHMARK
CMD is a publicly available collection of hundreds of thousands 2D maps and 3D grids containing different properties of the gas, dark matter, and stars from more than 2,000 different universes. The data has been generated from thousands of state-of-the-art (magneto-)hydrodynamic and gravity-only N-body simulations from the CAMELS project.
6 PAPERS • NO BENCHMARKS YET
This is a new dataset of news headlines and their frames related to the issue of gun violence in the United States. This Gun Violence Frame Corpus (GVFC) was curated and annotated by journalism and communication experts. The articles in this dataset are drawn from a sample of news articles from a list of 30 top U.S. news websites defined in terms of traffic to the websites; and collected from four time periods over the course of 2018 in order to capture a diversity of articles.
IQUAD is a dataset for Visual Question Answering in interactive environments. It is built upon AI2-THOR, a simulated photo-realistic environment of configurable indoor scenes with interactive object. IQUAD V1 has 75,000 questions, each paired with a unique scene configuration.
Imagewoof is a subset of 10 dog breed classes from Imagenet. The breeds are: Australian terrier, Border terrier, Samoyed, Beagle, Shih-Tzu, English foxhound, Rhodesian ridgeback, Dingo, Golden retriever, Old English sheepdog.
MARIDA (Marine Debris Archive) is the first dataset based on the multispectral Sentinel-2 (S2) satellite data, which distinguishes Marine Debris from various marine features that co-exist, including Sargassum macroalgae, Ships, Natural Organic Material, Waves, Wakes, Foam, dissimilar water types (i.e., Clear, Turbid Water, Sediment-Laden Water, Shallow Water), and Clouds. MARIDA is an open-access dataset which enables the research community to explore the spectral behaviour of certain floating materials, sea state features and water types, to develop and evaluate Marine Debris detection solutions based on artificial intelligence and deep learning architectures, as well as satellite pre-processing pipelines. Although it is designed to be beneficial for several machine learning tasks, it primarily aims to benchmark weakly supervised pixel-level semantic segmentation learning methods.
The MMBody dataset provides human body data with motion capture, GT mesh, Kinect RGBD, and millimeter wave sensor data. See homepage for more details.
Modern Office-31 is a refurbished version of the commonly used Office-31 dataset. Modern Office-31 rectifies many of the annotation errors and low quality images in the Amazon domain of the original Office-31 dataset. Additionally, this dataset adds another synthetic domain based on the Adaptiope dataset.
The REFLACX dataset contains eye-tracking data for 3,032 readings of chest x-rays by five radiologists. The dictated reports were transcribed and have timestamps synchronized with the eye-tracking data.
Understanding spatial relations (e.g., “laptop on table”) in visual input is important for both humans and robots. Existing datasets are insufficient as they lack largescale, high-quality 3D ground truth information, which is critical for learning spatial relations. In this paper, we fill this gap by constructing Rel3D: the first large-scale, human-annotated dataset for grounding spatial relations in 3D. Rel3D enables quantifying the effectiveness of 3D information in predicting spatial relations on large-scale human data. Moreover, we propose minimally contrastive data collection—a novel crowdsourcing method for reducing dataset bias. The 3D scenes in our dataset come in minimally contrastive pairs: two scenes in a pair are almost identical, but a spatial relation holds in one and fails in the other. We empirically validate that minimally contrastive examples can diagnose issues with current relation detection models as well as lead to sample-efficient training. Code and data are avai
SECOND is a well-annotated semantic change detection dataset. To ensure data diversity, we firstly collect 4662 pairs of aerial images from several platforms and sensors. These pairs of images are distributed over the cities such as Hangzhou, Chengdu, and Shanghai. Each image has size 512 x 512 and is annotated at the pixel level. The annotation of SECOND is carried out by an expert group of earth vision applications, which guarantees high label accuracy. For the change category in the SECOND dataset, we focus on 6 main land-cover classes, i.e. , non-vegetated ground surface, tree, low vegetation, water, buildings and playgrounds , that are frequently involved in natural and man-made geographical changes. It is worth noticing that, in the new dataset, non-vegetated ground surface ( n.v.g. surface for short) mainly corresponds to impervious surface and bare land. In summary, these 6 selected land-cover categories result in 30 common change categories (including non-change ). Through the
SNARE, short for ShapeNet Annotated with Referring Expressions, is a benchmark requires a model to choose which of two objects is being referenced by a natural language description.
This dataset aims at evaluating the License Plate Character Segmentation (LPCS) problem. The experimental results of the paper Benchmark for License Plate Character Segmentation were obtained using a dataset providing 101 on-track vehicles captured during the day. The video was recorded using a static camera in early 2015.
A multimodal dataset for sentiment analysis on internet memes.
Nearly 10,000 km² of free high-resolution and paired multi-temporal low-resolution satellite imagery of unique locations which ensure stratified representation of all types of land-use across the world: from agriculture to ice caps, from forests to multiple urbanization densities.
ARCADE: Automatic Region-based Coronary Artery Disease diagnostics using x-ray angiography imagEs Dataset Phase 2 consist of two folders with 300 images in each of them as well as annotations.
5 PAPERS • 2 BENCHMARKS
The complete blood count (CBC) dataset contains 360 blood smear images along with their annotation files splitting into Training, Testing, and Validation sets. The training folder contains 300 images with annotations. The testing and validation folder both contain 60 images with annotations. We have done some modifications over the original dataset to prepare this CBC dataset where some of the image annotation files contain very low red blood cells (RBCs) than actual and one annotation file does not include any RBC at all although the cell smear image contains RBCs. So, we clear up all the fallacious files and split the dataset into three parts. Among the 360 smear images, 300 blood cell images with annotations are used as the training set first, and then the rest of the 60 images with annotations are used as the testing set. Due to the shortage of data, a subset of the training set is used to prepare the validation set which contains 60 images with annotations.
5 PAPERS • NO BENCHMARKS YET
The dataset offers tag and mask annotations for image-text pairs from the CC3M validation set. Tag annotations denote words that aptly describe the relationship between the image and the corresponding text. These annotations provide valuable insights into the semantic connection between each pair's visual and textual elements.
The COCO-MIG benchmark (Common Objects in Context Multi-Instance Generation) is a benchmark used to evaluate the generation capability of generators on text containing multiple attributes of multi-instance objects. This benchmark consists of 800 sets of examples sampled from the COCO dataset. Following the layout of the COCO dataset, each instance is assigned random color information, and corresponding global image descriptions are constructed according to templates. The COCO-MIG also provides a complete pipeline for resampling and evaluating. For relevant tools and specific details, please refer to our project's homepage.
5 PAPERS • 1 BENCHMARK
ClueWeb22 is the newest iteration of the ClueWeb line of datasets, provides 10 billion web pages affiliated with rich information. Its design was influenced by the need for a high quality, large scale web corpus to support a range of academic and industry research, for example, in information systems, retrieval-augmented AI systems, and model pretraining. Compared with earlier CLUEWeb corpora, the ClUEWeb22 corpus is larger, more varied, of higher-quality, and aligned with the document distributions in commercial web search. Besides raw HTML, the dataset includes rich information about the web pages provided by industry-standard document understanding systems, including the visual representation of pages rendered by a web browser, parsed HTML structure information from a neural network parser, and pre-processed cleaned document text.
The DREAM dataset is introduce by the paper "Camera-to-Robot Pose Estimation from a Single Image" (ICRA 2020). This dataset consists of synthetic images (both with and without domain randomlization) of three different robot manipulators (Franka Emika’s Panda, Kuka’s LBR iiwa 7 R800, and Rethink Robotics’ Baxter) , as well as real-world images of Franka Emika’s Panda taken from various RGBD cameras (XBox 360 Kinect (XK), RealSense (RS), and Azure Kinect (AK)). Each instance in the dataset contains an RGB image, keypoint 3D/2D coordinates , global camera-to-robot transformation and joint state configurations (from both revolute and prismatic joint) of the robot. Tasks like estimating robot pose (camera pose) from a single RGB image, camera-to-robot calibration can be conducted and evaluated in this dataset.
Breast MRI scans of 922 cancer patients from Duke University, with tumor bounding box annotations, clinical, imaging, and many other features, and more.
DurLAR is a high-fidelity 128-channel 3D LiDAR dataset with panoramic ambient (near infrared) and reflectivity imagery for multi-modal autonomous driving applications. Compared to existing autonomous driving task datasets, DurLAR has the following novel features:
Approx. 300,000 images of galaxies labelled by shape.
Large language models (LLMs), after being aligned with vision models and integrated into vision-language models (VLMs), can bring impressive improvement in image reasoning tasks. This was shown by the recently released GPT-4V(ison), LLaVA-1.5, etc. However, the strong language prior in these SOTA LVLMs can be a double-edged sword: they may ignore the image context and solely rely on the (even contradictory) language prior for reasoning. In contrast, the vision modules in VLMs are weaker than LLMs and may result in misleading visual representations, which are then translated to confident mistakes by LLMs.
HowMany-Qa is a object counting dataset. It is taken from the counting-specific union of VQA 2.0 (Goyal et al., 2017) and Visual Genome QA (Krishna et al., 2016).
LabPics Chemistry Dataset
MIntRec is a novel dataset for multimodal intent recognition. It formulates coarse-grained and fine-grained intent taxonomies based on the data collected from the TV series Superstore. The dataset consists of 2,224 high-quality samples with text, video, and audio modalities and has multimodal annotations among twenty intent categories.
NewsCLIPpings is a dataset for detecting mismatched images and captions. Different to previous misinformation datasets, in NewsCLIPpings both the images and captions are unmanipulated, but some of them are mismatched.
The RAD-ChestCT dataset is a large medical imaging dataset developed by Duke MD/PhD Rachel Draelos during her Computer Science PhD supervised by Lawrence Carin. The full dataset includes 35,747 chest CT scans from 19,661 adult patients. The public Zenodo repository contains an initial release of 3,630 chest CT scans, approximately 10% of the dataset. This dataset is of significant interest to the machine learning and medical imaging research communities.
This dataset, called RodoSol-ALPR dataset, contains 20,000 images captured by static cameras located at pay tolls owned by the Rodovia do Sol (RodoSol) concessionaire, which operates 67.5 kilometers of a highway (ES-060) in the Brazilian state of Espírito Santo.
Satlas is a remote sensing dataset and benchmark that is large in both breadth, featuring all of the aforementioned applications and more, as well as scale, comprising 290M labels under 137 categories and 7 label modalities.