We introduce a large image dataset "Selfies and Video" for training a neural network to repel various attacks on biometric access systems. The dataset consists of a collection of selfies and videos recorded by individuals using both their front and web cameras. The main focus of the videos is on people pronouncing numbers in the wide range of individuals from various backgrounds and demographics.
"Selfies and Video Dataset" solves re-identification tasks and could be utilized in the field of facial recognition technology, for emotion detection and sentiment analysis applications, the dataset could also find application in virtual reality (VR) and augmented reality (AR) technologies. Overall, the dataset offers a wide range of applications in computer vision, artificial intelligence, social analysis, and human-computer interaction domains.
The dataset consists of 4,052 sets videos and selfies (32,416 media in total) from 3,500+ unique people from 15 countries. The data for the dataset is still gathering, so the number of videos and photos is getting bigger!
- Smartphone selfie - a selfie of a person made with a mobile phone, the person is depicted alone on it, the face is clearly visible.
- Smartphone video - filmed on the front camera, video of a person pronouncing numbers: duration of the first video is 30 seconds, duration of the second video is 8 seconds
- Webcam selfie - a selfie of a person made with a webcam, the person is depicted alone on it, the face is clearly visible.
- Webcam video - filmed on the webcam, video of a person pronouncing numbers: duration of the first video is 30 seconds, duration of the second video is 8 seconds
- 4 video files and 4 images made with phone front camera (2 videos and 2 photos) and web camera (2 videos and 2 photos).
- On video people pronounce numbers in Russian.
- People from 18 to 70 years old are presented in the dataset.
- For each person in the dataset age, country and gender is presented.
- The data was mostly collected indoor, however there are also selfies and videos made outdoors.
- The lighting is artificial, natural daily lightning, evening outdoor lighting and dark indoor lighting.
- People provided selfies and videos of themselves where the head takes up at least 1/2 of the frame.
- Distance from the camera is approximately 20-30 centimeters.
- corresponding to each person in the sample
- containing of selfies and videos of the individual
includes the following information for each media file:
- SetId: the identifier of the set,
- WorkerId: the identifier of the person who provided the set,
- Country: country of the person,
- Age: the age of the person,
- Gender: gender of the person,
- Type: type of the media file,
- Link: the link to access the media file
This is just an example of the data. If you need access to the entire dataset, contact us via sales@trainingdata.pro or leave a request on trainingdata.pro/data-market
| Resource | Link |
|---|---|
| TrainingData.pro | trainingdata.pro/data-market |
| Kaggle | https://www.kaggle.com/trainingdatapro |
| HuggingFace | https://huggingface.co/TrainingDataPro |