New

Please see an updated dataset with extended category list: Face Attributes Grouped

About

This dataset contains 4 categories of people with face accessories: plain, eyeglasses, sunglasses, and coverings. All the images were aligned and center-cropped to 256×256.

Note: the dataset may contain duplicates, imperfect alignment, or outlier images, not representative of the indicated category. There may also be mixed images, e.g., a picture of a person wearing spectacles and a mask.

Structure

There are a total of 5 directories: * plain [497] - images of regular people's faces * glasses [528] - images of people wearing spectacles * sunglasses [488] - images of people wearing sunglasses * sunglasses-imagenet [580] - additional images of people wearing sunglasses * covering [444] - images of people with covered faces (covered by hair, masks, hoodies, hands, face paint, etc.)

Collection

All the images were collected from Google Images using google-images-download with the usage rights flag labeled-for-noncommercial-reuse-with-modification. The exception is the sunglasses-imagenet folder which contains pictures from image-net, specifically downloaded from images.cv.

All the images were aligned with opencv-python and center-cropped to 256×256 with reflective padding. It is harder to align pictures with various face coverings, therefore, there are fewer images in that category.

License

This dataset is marked under CC BY-NC 4.0, meaning you can share and modify the data for non-commercial reuse as long as you provide a copyright notice. For the images downloaded from Google Images, you can find the original authors' licenses by looking up the image metadata based on their file names.

There are two folders in this Dataset. One is of people who are wearing glasses and one is of people who are not wearing glasses

2005-2016. This dataset includes data from the retired BRFSS Vision Module. From 2005-2011 the BRFSS employed a ten question vision module regarding vision impairment, access and utilization of eye care, and self-reported eye diseases. In 2013 and subsequently, one question in the core of BRFSS asks about vision: “Are you blind or do you have serious difficulty seeing, even when wearing glasses?” The latest data for this core question can be found in the Vision and Eye Health Surveillance System (VEHSS). VEHSS is intended to provide population estimates of vision loss function, eye diseases, health disparities, as well as barriers and facilitators to access to vision and eye care. This information can be used for designing, implementing, and evaluating vision and eye health prevention programs. To access the latest BRFSS data, (2013-2017) view the Behavioral Risk Factors – Vision and Eye Health Surveillance dataset (https://chronicdata.cdc.gov/Vision-Eye-Health/Behavioral-Risk-Factors-Vision-and-Eye-Health-Surv/vkwg-yswv).

About

The dataset contains synthetically generated images of people wearing glasses (regular eyeglasses + sunglasses) and glasses masks (full + frames + shadows). It can primarily be used for eyeglasses/sunglasses classification and segmentation.

This dataset is an augmented version of the synthetic dataset introduced in Portrait Eyeglasses and Shadow Removal by Leveraging 3D Synthetic Data and can be accessed here. The augmentation adds overlays on top of eyeglass frames to create images of people wearing sunglasses and corresponding masks.

Structure

There are 73 people identities in total, each with 400 different expressions or lightning effects, thus making a total of 29,000 samples. Each sample is a group of 8 images of the form sample-name-[suffix].png where [suffix] can be one of the following: * all - regular eyeglasses (i.e., frames) and their shadows * sunglasses - occluded glasses (i.e., sunglasses) and their frame shadows * glass - regular eyeglasses but no shadows * shadows - frame shadows but no eyeglasses * face - plain face: no glasses and no shadows * seg - mask for regular eyeglasses * sgseg - mask for sunglasses * shaseg - mask for frame shadows

10 identities were used for test data and 10 identities for validation, which corresponds to roughly 14% each, leaving around 72% of the data for training (which is 21,200 samples).

Collection

The data was generated in the following process: 1. The original dataset was downloaded from the link in the official Github repository 2. Glasses Detector was used to create full glasses segmentation masks which were used to generate various color and transparency (mainly dark) glasses 3. The generated glasses were overlaid on top of the original images with frames to create new images with sunglasses and corresponding masks 4. The 73 identities were shuffled and split into 3 parts (train, val, test) which were used to group all the 400 variations of each identity.

You can see the full process of glass overlay generation and data splitting in this gist.

Note: a type of noise (e.g., random, single spot) was added to roughly 15% of the images with sunglasses. Also, some of the generated glasses do not fill the entire frame, however, masks capture that.

Licence

This dataset is marked under CC BY-NC 4.0, meaning you can share and modify the data for non-commercial reuse as long as you provide a copyright notice.

Citation

Please use the original authors, i.e., the following citation:

@misc{glasses-segmentation-synthetic,
  author = {Junfeng Lyu, Zhibo Wang, Feng Xu},
  title = {Glasses Segmentation Synthetic Dataset},
  year = {2023},
  publisher = {Kaggle},
  journal = {Kaggle datasets},
  howpublished = {\url{https://www.kaggle.com/datasets/mantasu/glasses-segmentation-synthetic-dataset}}
}

93604 images of faces with or without glasses. Faces are cropped using Mediapipe's face detector (blaze face).

image format .jpg

image sizes 256x256 with 3 channels

test - 3803 test\without_glasses test\with_glasses

train - 76049 without_glasses train\with_glasses

val - 15210 val\without_glasses val\with_glasses

classification criteria

with_glasses: people with eyeglasses, sunglasses, swimming/safety goggles, etc. on face, on nose, or on eyes with_glasses: people without any kind of glasses, but perhaps with face mask, face painting, eye patches etc.

The 110 People – Human Face Image Data is gathered through camera shot involving 110 participants, with a proper balance of gender ratio and age group distribution covering major skin tones. Each person contributes 2100 pictures with glasses/ no glasses, expressions, camera shooting angle, and lighting conditions. All Attributes are annotated such as gender, age, expression, etc. The overall accuracy rate is ≥ 97%.This dataset is suitable for face recognition, facial expression analysis, and AI training.

Introduction

Welcome to the Native American Human Face with Occlusion Dataset, carefully curated to support the development of robust facial recognition systems, occlusion detection models, biometric identification technologies, and KYC verification tools. This dataset provides real-world variability by including facial images with common occlusions, helping AI models perform reliably under challenging conditions.

Facial Image Data

The dataset comprises over 3,000 high-quality facial images, organized into participant-wise sets. Each set includes:

Diversity & Representation

Image Quality & Capture Conditions

To ensure robustness and real-world utility, images were captured under diverse conditions:

Metadata

Each image is paired with detailed metadata to enable advanced filtering, model tuning, and analysis:

This rich metadata helps train models that can recognize faces even when partially obscured.

Use Cases & Applications

This dataset is ideal for a wide range of real-world and research-focused applications, including:

Secure & Ethical Collection

Dataset Updates &

Frequency of myopia (all, childhood-onset and adult-onset) and emmetropia in the UK Biobank population: Distribution of socio-demographic and environmental factors.

Introduction

Welcome to the African Human Face with Occlusion Dataset, carefully curated to support the development of robust facial recognition systems, occlusion detection models, biometric identification technologies, and KYC verification tools. This dataset provides real-world variability by including facial images with common occlusions, helping AI models perform reliably under challenging conditions.

Facial Image Data

The dataset comprises over 5,000 high-quality facial images, organized into participant-wise sets. Each set includes:

Diversity & Representation

Image Quality & Capture Conditions

To ensure robustness and real-world utility, images were captured under diverse conditions:

Metadata

Each image is paired with detailed metadata to enable advanced filtering, model tuning, and analysis:

This rich metadata helps train models that can recognize faces even when partially obscured.

Use Cases & Applications

This dataset is ideal for a wide range of real-world and research-focused applications, including:

Secure & Ethical Collection

Dataset

Dataset

This dataset was created by Cap_Apoorv

Contents

eSight designs and manufactures electronic eyewear for people with severe vision loss.

Full edition for scientific use. More than 37 million handicapped people live in the European Union. The Council of Europe has declared the year 2003 the European year of handicapped people to give them a chance to make people all over Europe aware of their interests. One aim is the development of the measures needed to better integrate handicapped persons in all parts of social life. In this context, the European Commission consigned a EU-wide study (performed as an ad-hoc module to the LFS) on the employment of handicapped people in the year 2002. The goal was to collect a comprehensive and connected data set on the employments situation of handicapped people. In Austria, the study was conducted as a Microcensus special survey (“handicaps and disabilities”) in June 2002. Starting point of the survey program was the question on lasting health problems or handicaps. The term “lasting” denoted a time span of at least 6 months. The basic concept of health problems is broad. It includes physical injury, sensory problems (e.g. subjectively felt visual impairments despite wearing glasses), heart- and breathing troubles and walking impairment, as well as other progressive diseases (e.g. cancer, Parkinson disease, etc.), psychological problems and learning disabilities. In addition to the question on the existence of a lasting health problem or a handicap, it was asked whether this problem or handicap affected the everyday life. “Everyday life” denotes important personal performances, for instance eating, washing/bathing, exercises such as climbing the stairs, going shopping, cooking or doing the laundry. An impairment of theses performances exists if they persist despite the usage of aids (e.g. hearing aid, glasses), the utilisation of medical treatments and the help of other people.

Weighted crude (univariable) associations with uncorrected refractive error (URE) and need for reading glasses.

Introduction

Welcome to the Middle Eastern Human Face with Occlusion Dataset, carefully curated to support the development of robust facial recognition systems, occlusion detection models, biometric identification technologies, and KYC verification tools. This dataset provides real-world variability by including facial images with common occlusions, helping AI models perform reliably under challenging conditions.

Facial Image Data

The dataset comprises over 3,000 high-quality facial images, organized into participant-wise sets. Each set includes:

Diversity & Representation

Image Quality & Capture Conditions

To ensure robustness and real-world utility, images were captured under diverse conditions:

Metadata

Each image is paired with detailed metadata to enable advanced filtering, model tuning, and analysis:

This rich metadata helps train models that can recognize faces even when partially obscured.

Use Cases & Applications

This dataset is ideal for a wide range of real-world and research-focused applications, including:

Secure & Ethical Collection

Dataset

This dataset contains structural, thermophysical, and mechanical properties of SiO₂-B₂O₃-Na₂O glasses obtained from Molecular Dynamics simulations. This includes the population of the different structural glass-forming units, glass densities, enthalpies of mixing, elastic moduli (Young, bulk, and shear), and Poisson's ratio. The data cover the whole glass-forming composition space of the SiO₂-B₂O₃-Na₂O pseudo-ternary system with a constant composition increment of 5%. Two different empirical potentials were used in the simulations to assess the potential-dependent character of the observations. The dataset includes experimental values for the density and mechanical properties of several glass compositions gathered from the literature, which were used to quantify the validity of the two potentials.

Dataset

This dataset was created by Coffeeshop

Released under CC BY-NC-SA 4.0

Contents

Dataset

This dataset was created by AHMAD TALHA ANSARI

Released under MIT

Contents

THÖR is a dataset with human motion trajectory and eye gaze data collected in an indoor environment with accurate ground truth for the position, head orientation, gaze direction, social grouping and goals. THÖR contains sensor data collected by a 3D lidar sensor and involves a mobile robot navigating the space. In comparison to other, our dataset has a larger variety in human motion behaviour, is less noisy, and contains annotations at higher frequencies.

THÖR eye-tracking - data of the participant (Helmet number 9) from the Tobii Glasses included in this dataset. The entire data from the experiment was recorded into two recordings - "Recording011" and "Recording012".

The folder "RawData" consists of exported data from Tobii Pro Lab software using a filter called "Tobii-IVT Attention filter" (velocity threshold parameter set to 100 degrees/second), which is a recommended method for dynamic situations. The recommendation was given by the equipment manufactures Tobii Pro and from other researchers. For further information, please refer to https://www.tobiipro.com/siteassets/tobii-pro/user-manuals/Tobii-Pro-Lab-User-Manual/?v=1.86 (Appendix-B, page 85).

The recording start times are as follows:
Tobii recording011 starttime: 13:34:37.267
Tobii recording012 starttime: 14:36:17.730

The folder "SynchronisedData" of "ALL_Qualisys_Tobii.mat" file, which consists of all the synchronised data between Qualisys data and Tobii eye-tracker data using timestamps matching. The columns (headers) in this mat file respectively represent - 'Timestamp', 'Pos_X', 'Pos_Y', 'Pos_Z', 'Head_R', 'Head_P', 'Head_Y','GazepointX', 'GazepointY', 'Gaze3DpositioncombinedX', 'Gaze3DpositioncombinedY', 'Gaze3DpositioncombinedZ', 'Gaze3DpositionleftX', 'Gaze3DpositionleftY', 'Gaze3DpositionleftZ', 'Gaze3DpositionrightX', 'Gaze3DpositionrightY', 'Gaze3DpositionrightZ', 'GazedirectionleftX', 'GazedirectionleftY', 'GazedirectionleftZ', 'GazedirectionrightX', 'GazedirectionrightY', 'GazedirectionrightZ', 'PupilpositionleftX', 'PupilpositionleftY', 'PupilpositionleftZ', 'PupilpositionrightX', 'PupilpositionrightY', 'PupilpositionrightZ', 'Pupildiameterleft', 'Pupildiameterright', 'Recordingmedianame', 'Recordingmediawidth', 'Recordingmediaheight', 'Gazeeventduration', 'Eyemovementtypeindex', 'FixationpointX', 'FixationpointY', 'GyroX', 'GyroY', 'GyroZ', 'AccelerometerX', 'AccelerometerY', 'AccelerometerZ'.

'Timestamp', 'Pos_X', 'Pos_Y', 'Pos_Z', 'Head_R', 'Head_P', 'Head_Y','GazepointX', 'GazepointY' represent the timestamps (matched to Tobii timestamps using nearest neighbor search), position and head orientation from Qualisys data and rest of the data is from Tobii eye-tracker. For more information regarding the eye-tracker data, please refer to https://www.tobiipro.com/siteassets/tobii-pro/user-manuals/Tobii-Pro-Lab-User-Manual/?v=1.86 (Section 8.7.2.1, page 68).

Visual problems are common in people who have neurological injury or disease, with deficits linked to postural control and gait impairment. Vision therapy could be a useful intervention for visual impairment in various neurological conditions such as stroke, head injury, or Parkinson’s disease. Stroboscopic visual training (SVT) has been shown to improve aspects of visuomotor and cognitive performance in healthy populations, but approaches vary with respect to testing protocols, populations, and outcomes. The purpose of this structured review was to examine the use of strobe glasses as a training intervention to inform the development of robust protocols for use in clinical practice. Within this review, any studies using strobe glasses as a training intervention with visual or motor performance–related outcomes was considered. PubMed, Scopus, and ProQuest databases were searched in January 2023. Two independent reviewers (JD and RM) screened articles that used strobe glasses as a training tool. A total of 33 full text articles were screened, and 15 met inclusion/exclusion criteria. Reported outcomes of SVT included improvements in short–term memory, attention, and visual response times, with emerging evidence for training effects translating to balance and physical performance. However, the lack of standardisation across studies for SVT protocols, variation in intervention settings, duration and outcomes, and the limited evidence within clinical populations demonstrates that further work is required to determine optimal strobe dosage and delivery. This review highlights the potential benefits, and existing research gaps regarding the use of SVT in clinical practice, with recommendations for clinicians considering adopting this technology as part of future studies in this emerging field.

93604 images of people with or without glasses

image format .jpg

image sizes 500x500 with 3 channels

test - 3745 test\without_glasses - 2380 test\with_glasses - 1365

train - 74883 without_glasses - 47586 train\with_glasses - 27297

val - 14976 val\without_glasses - 9517 val\with_glasses - 5459

classification criteria

with_glasses: people with eyeglasses, sunglasses, swimming/safety goggles, etc. on face, on nose, or on eyes with_glasses: people without any kind of glasses, but perhaps with face mask, face painting, eye patches etc.