Papers with Code 아카이브

2D Classification

태스크의 하위 분류와 논문·코드·평가·기법·데이터셋을 탐색하세요.

하위 태스크
27
평가표
0
보존된 평가 기록
0

데이터셋

출처에 같은 태스크 이름으로 연결된 데이터셋입니다. 동명 태스크가 포함될 수 있으므로, 비교할 때는 평가 결과의 조건도 확인하세요.

SSBI Dataset

The Synthetic Signature Bankcheck Images (SSBI) Dataset is the first publicly available dataset of bank check images with annotations for detecting handwritten components, including names, amounts, dates, and signatures. It also supports both writer-independent and writer-dependent signature verification tasks by providing labels for genuine and forged signatures and IDs of the signature authors. This dataset contains 2180 synthetic bank check images with: -1680 genuine signatures - 500 forged signatures The signatures are from 19 different signers. In addition to the signature bounding boxes, the dataset also contains the bounding boxes for the date, amount, and payee fields.

SSL

This is a dataset to benchmark real-time embedded object detection models for RoboCup SSL (Small Size League).

STN PLAD

STN PLAD is a high-resolution and real-world image dataset of multiple high-voltage power line components. It has 2,409 annotated objects divided into five classes: transmission tower, insulator, spacer, tower plate, and Stockbridge damper, which vary in size (resolution), orientation, illumination, angulation, and background. ## Properties - Image size: 5472×3078 or 5472×3648 - Total images: 133 - Total instances: 2409 - Average instances per image: 18.1 - Nº of object classes (different assets): 5 - Other stats: ## ## Abstract Many power line companies are using UAVs to perform their inspection processes instead of putting their workers at risk by making them climb high voltage power line towers, for instance. A crucial task for the inspection is to detect and classify assets in the power transmission lines. However, public data related to power line assets are scarce, preventing a faster evolution of this area. This work proposes the Power Line Assets Dataset, containing high-resolution and real-world images of multiple high-voltage power line components. It has 2,409 annotated objects divided into five classes: transmission tower, insulator, spacer, tower plate, and Stockbridge damper, which vary in size (resolution), orientation, illumination, angulation, and background. This work also presents an evaluation with popular deep object detection methods, showing considerable room for improvement. ## Baseline results - mAP: 89.2% | Assets | Average Precision | |--------------------|-----------------| | Transmission tower | 0.900 | | Insulator | 0.894 | | Spacer | 0.856 | | Tower plate | 0.971 | | Stockbridge damper | 0.838 | | mean | 0.892 |

Street Dataset

A real-world image dataset that contains more than 900 images generated from 26 street cameras and 7 object categories annotated with detailed bounding box. The data distribution is non-IID and unbalanced, reflecting the characteristic real-world federated learning scenarios. Source: Real-World Image Datasets for Federated Learning

SunspotsYoloDataset: annotated solar images captured with smart telescopes (January 2023 - May 2024)

SunspotsYoloDataset is a set of 1690+380+128 high-resolution RGB astronomical images captured with smart telescopes with specific solar filters and annotated with the positions of sunspots that are effectively in the images. Two instruments were used for several months from Luxembourg and France between January 2023 and May 2024: a Stellina smart telescope (https://vaonis.com/stellina) and a Vespera smart telescope (https://vaonis.com/vespera). SunspotsYoloDataset can be used to train YOLO detection models on solar images, enabling the prediction of unexpected events such as Borealis Aurora with astronomical equipment accessible to the public. SunspotsYoloDataset is formatted with the YOLO standard, i.e., with separated files for images and annotations, usable by state-of-the-art training tools and graphical software like MakeSense (https://www.makesense.ai). More precisely, there is a ZIP file containing RGB images in JPEG format (minimal compression), and text files containing the positions of sunspots. Each RGB image has a resolution of 640 × 640 pixels. For more details about the dataset, please contact the author: olivier.parisot@list.lu . For more information about Luxembourg of Science and Technology (LIST), please consult: https://www.list.lu .

SyntheticFur

SyntheticFur is a dataset for neural rendering. Collecting and generating high quality fur images is an expensive and difficult process that requires content specialists to generate. By releasing this unique dataset with high quality lighting simulation via ray tracing, this can save time for researchers seeking to advance studies of fur rendering and simulation, without having to recreate this laborious process. The dataset was used for neural rendering research at Google that takes advantage of rasterized image buffers and converts them into high quality raytraced fur renders. We believe that this dataset can contribute to the computer graphics and machine learning community to develop more advanced techniques with fur rendering. It contains approximately 140,000 procedurally generated images and 15 simulations with Houdini. The images consist of fur groomed with different skin primitives and move with various motions in a predefined set of lighting environments.

TAMPAR

TAMPAR is a real-world dataset of parcel photos for tampering detection with annotations in COCO format. For details see the paper and for visual samples the project page. Features are: - >900 annotated real-world images with >2,700 visible parcel side surfaces - 6 different tampering types - 6 different distortion strengths

Tea sickness - object detection

Please refer this paper @article{sarker2024tea, author = {Sarker, Swapnil Sharma and Islam, Ashiqul and Talukder Raktim, Raufun and Roshni, Sanjana and Joy, Sajib Kumar Saha and Shah, Faisal}, title = {Real-Time Tea Leaf Disease Detection Using Deep Learning-Based Models}, journal = {Journal Name or Preprint Platform}, year = {2024}, month = {November}, doi = {10.13140/RG.2.2.10031.24483} }

ThermoScenes

Dataset of paired thermal and RGB images comprising ten diverse scenes—six indoor and four outdoor scenes— for 3D scene reconstruction and novel view synthesis (e.g. with NeRF).

THGP

Temporal Hands Guns and Phones (THGP) dataset, is a collection of 5960 video frames (5000 for training and 960 for testing). The training part is composed with 50 videos of 100 frames (720 × 720 pixels). This dataset contains 20 videos of shooting drills, 20 videos of armed robberies, and 10 videos of people making calls. The testing part contains 48 videos of 20 frames (720 × 720). Videos contained in the testing dataset includes phone calls, gun reviews, shooting drills, people making calls, and armed robberies at convenience stores. This dataset is labeled with the bounding boxes of hands, phones, and guns.

Three-view Synthetic data

10000 instances of three-view numerical data set with 4 clusters and 2 feature components are considered. The data points in each view are generated from a 2-component 2-variate Gaussian mixture model (GMM) where their mixing proportions $\alpha_1^{(1)}=\alpha_1^{(2)}=\alpha_1^{(3)}=\alpha_1^{(4)}=0.3$; $\alpha_2^{(1)}=\alpha_2^{(2)}=\alpha_2^{(3)}=\alpha_2^{(4)}=0.15$; $\alpha_3^{(1)}=\alpha_3^{(2)}=\alpha_3^{(3)}=\alpha_3^{(4)}=0.15$ and $\alpha_4^{(1)}=\alpha_4^{(2)}=\alpha_4^{(3)}=\alpha_4^{(4)}=0.4$. The means $\mu_{ik}^{(1)}$ for the first view are $[-10 ~-5)]$,$[-9 ~ 11]$, $[0~ 6]$ and $[4~0]$; The means $\mu_{ik}^{(2)}$ for the view 2 are $[-8 ~-12]$,$[-6 ~ -3]$, $[-2~ 7]$ and $[2~1]$; And the means $\mu_{ik}^{(3)}$ for the third view are $[-5 ~-10]$,$[-8 ~ -1]$, $[0~ 5]$ and $[5~-4]$. The covariance matrices for the three views are $\Sigma_1^{(1)}=\Sigma_1^{(2)}=\Sigma_1^{(3)}=\Sigma_1^{(4)}=\left[ \begin{array}{cc} 1 & 0\\0&1\end{array}\right]$; $\Sigma_2^{(1)}=\Sigma_2^{(2)}=\Sigma_2^{(3)}=\Sigma_2^{(4)}=3 \left[ \begin{array}{cc} 1 & 0\\0&1\end{array}\right]$; $\Sigma_3^{(1)}=\Sigma_3^{(2)}=\Sigma_3^{(3)}=\Sigma_3^{(4)}=2 \left[ \begin{array}{cc} 1 & 0\\0&1\end{array}\right]$; and $\Sigma_4^{(1)}=\Sigma_4^{(2)}=\Sigma_4^{(3)}=\Sigma_4^{(4)}=0.5 \left[ \begin{array}{cc} 1 & 0\\0&1\end{array}\right]$. These $x_1^{(1)}$ and $x_2^{(1)}$ are the coordinates for the view 1, $x_1^{(2)}$ and $x_2^{(2)}$ are the coordinates for the view 2, $x_1^{(3)}$ and $x_2^{(3)}$ are the coordinates for the view 3. While the original distribution of data points for cluster 1, cluster 2, cluster 3, and cluster 4 are 1514, 3046, 3903, and 1537, respectively.

TimberVision

The TimberVision dataset consists of more than 2k annotated RGB images and contains a total of 51k trunk components including cut and lateral surfaces, thereby surpassing any existing dataset in this domain in terms of both quantity and detail by a large margin. The dataset can be used to train oriented object detection and instance segmentation and evaluate the influence of multiple scene parameters on model performance. Additionally, a generic framework is provided to fuse the components detected by the models for both tasks into unified trunk representations. Furthermore, geometric properties are derived automatically and multi-object tracking is applied to further enhance robustness.

TiROD

Dataset to benchmark Continual Learning for Object Detection in a Tiny Robotics settings. * 10 Continual learning tasks * 5 environments (1 indoor and 4 outdoor) * 2 illumination conditions * 14 object categories Dataset Website Dataset video Code Paper

TPIC17

Image dataset with about 600K Flickr photos. Source: Sequential Prediction of Social Media Popularity with Deep Temporal Context Networks

UAVBillboards

Mapping urban large-area advertising structures using drone imagery and deep learning-based spatial data analysis. The dataset contains 1361 images supplemented with additional spatial metadata, together with 5210 annotations in a COCO-like format.

UAVVaste

The UAVVaste dataset consists to date of 772 images and 3716 annotations. The main motivation for creation of the dataset was the lack of domain-specific data. The datasets that are widely used for object detection evaluation benchmarking. The dataset is made publicly available and is intended to be expanded. | Date | Images count | Annotations count | |------------ |:------------: |:-----------------: | | 14.11.2020 | 772 | 3716 |

uBench

Microscopy is a cornerstone of biomedical research, enabling detailed study of biological structures at multiple scales. Advances in cryo-electron microscopy, high-throughput fluorescence microscopy, and whole-slide imaging allow the rapid generation of terabytes of image data, which are essential for fields such as cell biology, biomedical research, and pathology. These data span multiple scales, allowing researchers to examine atomic/molecular, subcellular/cellular, and cell/tissue-level structures with high precision. A crucial first step in microscopy analysis is interpreting and reasoning about the significance of image findings. This requires domain expertise and comprehensive knowledge of biology, normal/abnormal states, and the capabilities and limitations of microscopy techniques. Vision-language models (VLMs) offer a promising solution for large-scale biological image analysis, enhancing researchers’ efficiency, identifying new image biomarkers, and accelerating hypothesis generation and scientific discovery. However, there is a lack of standardized, diverse, and large-scale vision-language benchmarks to evaluate VLMs’ perception and cognition capabilities in biological image understanding. To address this gap, we introduce μ-Bench, an expert-curated benchmark encompassing 22 biomedical tasks across various scientific disciplines (biology, pathology), microscopy modalities (electron, fluorescence, light), scales (subcellular, cellular, tissue), and organisms in both normal and abnormal states. We evaluate state-of-the-art biomedical, pathology, and general VLMs on μ-Bench and find that: i) current models struggle in all categories, even for basic tasks such as distinguishing microscopy modalities; ii) current specialist models fine-tuned on biomedical data often perform worse than generalist models; iii) fine-tuning in specific microscopy domains can cause catastrophic forgetting, eroding prior biomedical knowledge encoded in their base model. iv) weight interpolation between fine-tuned and pre-trained models offers one solution to forgetting and improves general performance across biomedical tasks. We release μ-Bench under a permissive license to accelerate the research and development of microscopy foundation models.

UCCS

Unconstrained Face Detection and Open-Set Face Recognition Challenge * Paper: https://arxiv.org/abs/1708.02337 * Official website: https://vast.uccs.edu/Opensetface * UnOfficial website: https://exposing.ai/uccs Face detection and recognition benchmarks have shifted toward more difficult environments. The challenge presented in this paper addresses the next step in the direction of automatic detection and identification of people from outdoor surveillance cameras. While face detection has shown remarkable success in images collected from the web, surveillance cameras include more diverse occlusions, poses, weather conditions and image blur. Although face verification or closed-set face identification have surpassed human capabilities on some datasets, open-set identification is much more complex as it needs to reject both unknown identities and false accepts from the face detector. We show that unconstrained face detection can approach high detection rates albeit with moderate false accept rates. By contrast, open-set face recognition is currently weak and requires much more attention.

UG^2

Contains three difficult real-world scenarios: uncontrolled videos taken by UAVs and manned gliders, as well as controlled videos taken on the ground. Over 160,000 annotated frames forhundreds of ImageNet classes are available, which are used for baseline experiments that assess the impact of known and unknown image artifacts and other conditions on common deep learning-based object classification approaches. Source: UG^2: a Video Benchmark for Assessing the Impact of Image Restoration and Enhancement on Automatic Visual Recognition

USB

The Universal-Scale object detection Benchmark (USB) is a benchmark for object detection that has variations in object scales and image domains by incorporating COCO with the recently proposed Waymo Open Dataset and Manga109-s dataset. To enable fair comparison, USB establishes different protocols by defining multiple thresholds for training epochs and evaluation image resolutions.

USC-GRAD-STDdb

USC-GRAD-STDdb comprises 115 video segments containing more than 25,000 annotated frames of HD 720p resolution (≈1280x720) with small objects of interest from 16 (≈4x4) to 256 (≈16x16) as pixel area. The length of the videos changes from 150 up to 500 frames. The size of every object is determined through the bounding box, so that a good annotation is of utmost importance for reliable performance metrics. As it may seem obvious, the smaller the object, the harder the annotation. The annotation has been carried out with the ViTBAT tool, adjusting the boxes as much as possible to the objects of interest in each video frame. In total, more than 56,000 ground truth labels have been generated.

VESUS

The Varied Emotion in Syntactically Uniform Speech (VESUS) repository is a lexically controlled database collected by the NSA lab. Here, actors read a semantically neutral script of words, phrases, and sentences with different emotional inflections. VESUS contains 252 distinct phrases, each read by 10 actors in 5 emotional states (neutral, angry, happy, sad, fearful). Source: https://engineering.jhu.edu/nsa/vesus/

VETRA

VETRA is a dataset for vehicle tracking in aerial image sequences and presents unique challenges such as low frame rates, small and fast-moving objects, as well as high camera movement. These characteristics allow for extended tracking of numerous vehicles with varying motion behaviors over large areas and pose new challenges for MOT algorithms. VETRA consists of 52 image sequences captured by airplanes and helicopters using DLR’s 3k and 4k camera systems. The acquisition sites are located in Germany and Austria. In addition to the classical training, validation and test sets, VETRA offers a second test set specifically designed for the application of large area monitoring (LAM). The LAM sequences are recorded over 7 rural roads and motorways with a fixed camera speed and configuration. Each road section is captured at 4 different times of the day, enabling the performance of MOT algorithms to be evaluated under different traffic loads in a static environment. Furthermore, the features extracted from the LAM sequences can be utilized in transport research applications.

VGGFace2 HQ

A high-resolution version of VGGFace2 for academic face editing purposes. This project uses GFPGAN for image restoration and insightface for data preprocessing (crop and align).

Vinegar Fly

Vinegar Fly is a pose estimation dataset for fruit flies.

VisA-AC

VisA-AC is a refined benchmark based on the VisA dataset, tailored for the task of anomaly classification—distinguishing between different types of anomalies rather than simply detecting whether an image is anomalous. While the original VisA provides anomaly type information in an Excel file, it includes numerous under-sampled and ambiguous classes. VisA-AC addresses these issues by removing classes with fewer than 10 samples, merging visually similar categories, and manually correcting mislabeled samples. Additionally, anomaly classes in VisA-AC are organized into separate folders—following the structure of MVTec-AC—for easier integration and usage. The resulting dataset ensures both statistical robustness and semantic clarity, supporting rigorous evaluation of multi-class anomaly classification methods in real-world industrial settings.

VISEM-Tracking

VISEM-Tracking is a dataset consisting of 20 video recordings of 30s of spermatozoa with manually annotated bounding-box coordinates and a set of sperm characteristics analyzed by experts in the domain. It is an extension of the previously published VISEM dataset. In addition to the annotated data, unlabeled video clips are provided for easy-to-use access and analysis of the data. Source: VISEM-Tracking: Human Spermatozoa Tracking Dataset Image Source: https://arxiv.org/pdf/2212.02842v1.pdf

Visual 3D shape matching dataset

Dataset for testing the ability of Vision Language Models (LVM) to recognize and match 3D objects of the exact same 3D shapes but with different orientation/materials/textures/ environments and light conditions.

VizWiz-FewShot

VizWiz-FewShot is a a few-shot localization dataset originating from photographers who authentically were trying to learn about the visual content in the images they took. It includes nearly 10,000 segmentations of 100 categories in over 4,500 images that were taken by people with visual impairments.

VME & CDSI

Vehicles in the Middle East (VME) dataset, designed explicitly for vehicle detection in high-resolution satellite images from Middle Eastern countries. Sourced from Maxar, the VME dataset spans 54 cities across 12 countries, comprising over 4,000 image tiles and more than 100,000 vehicles, annotated using both manual and semi-automated methods. Also, we introduce the largest benchmark dataset for Car Detection in Satellite Imagery (CDSI), combining images from multiple sources to enhance global car detection.

VQA-OV

Collects 60 reference sequences and 540 impaired sequences. Source: Bridge the Gap Between VQA and Human Behavior on Omnidirectional Video: A Large-Scale Dataset and a Deep Learning Model

YorkTag

YorkTag provides pairs of sharp/blurred images containing fiducial markers and is proposed to train and qualitatively and quantitatively evaluate our model.

YouTube-GDD

YouTubeGun Detection Dataset is collected from 343 high-definition YouTube videos and contains 5000 well-chosen images, in which 16064 instances of gun and 9046 instances of person are annotated. Compared to other datasets, YouTube-GDD is "dynamic", containing rich contextual information

AppleBBCH76

The photo fixation of apple fruitlets was done in the LatHort orchard in Dobele, at the development of fruit (BBCH stage 76-78). BBCH-scale describes the phenological development of grapes: 7 - development of fruit; 76 - fruit about 60% final size; 78 - fruit about 80% final size. Two photo images were taken for each tree – perpendicularly, in a tree-facing view and in an oblique view. The images were annotated using the tool makesense.ai. Then the annotated images 3008x2000 were automatically cropped out on 640x640 images with overlap 30% and validated manually. The images were saved in YOLO format. Source: <a href="https://www.kaggle.com/datasets/projectlzp201910094/applebbch76">https://www.kaggle.com/datasets/projectlzp201910094/applebbch76</a> Dataset was presented in the article: Kodors, S., Zarembo, I., Lācis, G., Litavniece, L., Apeināns, I., Sondors, M., & Pacejs, A. (2024). Autonomous Yield Estimation System for Small Commercial Orchards Using UAV and AI. Drones, 8(12), 734. <a href="https://doi.org/10.3390/drones8120734">https://doi.org/10.3390/drones8120734</a>

AppleBBCH81

The photo fixation of apple fruits was done in the LatHort orchard in Dobele, at the maturity of fruit and seed (BBCH stage 81-85). BBCH-scale describes the phenological development of grapes: 8 - maturity of fruit and seed; 81 - beginning of ripening; 85 - advanced ripening. Two photo images were taken for each tree – perpendicularly, in a tree-facing view and in an oblique view. The images were annotated using the tool makesense.ai. Then the annotated images 3008x2000 were automatically cropped out on 640x640 images with overlap 30% and validated manually. The images were saved in YOLO format. Source: <a href="https://www.kaggle.com/datasets/projectlzp201910094/applebbch81">https://www.kaggle.com/datasets/projectlzp201910094/applebbch81</a> Dataset was presented in the article: Kodors, S., Zarembo, I., Lācis, G., Litavniece, L., Apeināns, I., Sondors, M., & Pacejs, A. (2024). Autonomous Yield Estimation System for Small Commercial Orchards Using UAV and AI. Drones, 8(12), 734. <a href="https://doi.org/10.3390/drones8120734">https://doi.org/10.3390/drones8120734</a>

출처: Papers with Code · CC BY-SA 4.0. 아카이브 자료와 새로 수집한 논문은 별도로 관리합니다.