INTEL-TAU

Kuvaus

__Introduction__ This dataset, called INTEL-TAU is a set of 7022 scenes in total, which makes it the largest available dataset for illumination estimation research. The variety of scenes captured using three different camera models, i.e., Canon 5DSR, Nikon D810, and Sony IMX135, makes the dataset appropriate for evaluating the camera and scene invariance of the different illumination estimation techniques. Privacy masking is done for sensitive information, e.g., faces. Thus, it is coherent with the General Data Protection Regulation(GDPR) regulations. Furthermore, the effect of color shading for mobile images can be evaluated with INTEL-TAU, as we provide both corrected and uncorrected versions of the raw data. The main variant of the dataset is saved in Canon_5DSR_1080p.zip, Nikon_D180_1080p.zip, and Sony_IMX135_BLCCSC_1080p.zip. This variant is suitable for color constancy research. The black level was subtracted from all the scenes and the saturation points were normalized and all images were resized to 1080p. __Organization of the Dataset__ The images in the dataset are organized as follows: 'RootPath/CameraName/SetName'. -CameraName: Indicates the name of the camera used to take the image, and it is one of the following: 1) Canon_5DSR 2) Nikon_D180 3) Sony_IMX135 4) Sony_IMX135_BLCCSC: Sony_IMX135 images after color shading correction. -SetName: Indicates the name of the set (category) of images and can take following names: 1) field_1_cameras: scenes unique for the camera model. 2) field_3_cameras: common scenes acquired by the three cameras. 3) lab_printouts: lab printouts. 4) lab_realscene: lab real scenes. The contents are as follows for each image, and follow the above annotation with corresponding extensions: 1) Ground truth: This file contains the white point in normalized [R, G, B] format. Extension: '.wp'. 2) TIFF Image: These images underwent black level correction, saturation point stretching operations, and are downscaled to 1080p. Note: splits.zip contains the proposed non-random 10-fold split of the data to be used for cross-validation of color constancy approaches. __Additional Resources__ We also provide the original raw version, i.e., raw Bayer, of the dataset ( Canon_5DSR.zip, Nikon_D180.zip, Sony_IMX135.zip, Sony_IMX135_BLCCSC.zip). The white points are stored as [R/G, B/G] coordinates. The spectral responses of the different camera models and the spectral power distributions of the lab light sources are also provided. It should be noted that the size of the raw image set is large, 290GB, compared to 50GB of the preprocessed dataset images. This variant of the dataset can be used to study the color shading effect as we provide the mobile images before and after color shading. Since raw Bayer images are provided, this dataset can also be used to study various imaging problems, such as demosaicing. The contents are as follows for each image, and follow the above annotation with corresponding extensions: 1) Color Correction Matrix: The color correction matrix from sensor RGB to sRGB. It is stored as a 1x9 vector. Extension: '.ccm'. 2) JPEG Image: These images underwent demosaicing, black level correction, saturation point stretching operations, white balancing, color correction, and gamma correction. These images are only shared for illustration of how a color corrected image should look like. Extension: '.jpg'. 3) Ground truth: This file contains the white point in [R/G, B/G] format. Extension: '.wp'. 4) RAW Image: 2 bytes per pixel raw format, storing the actual raw Bayer layout, the values can be read as uint16. Extension: '.plain16'. Note: The 'RootPath/CameraName/Info/Info/CameraName_frame_format_ .mat files contain the width and height information, as well as the Bayer order and other needed frame format information needed to correctly read the raw image. The additional folder 'RootPath/Info' has additional information: Spectral Light Source Data and Spectral Responses of the Cameras. Isotropic variant: We also provide a preprocessed downscaled version of the dataset with isotropic scaling: in this variant, the resized images have a height of 1080 lines and the width was adjusted accordingly to preserve the original aspect ratio of the images. This variant is saved in Canon_5DSR_isotropic.zip, Nikon_D180_isotropic.zip, and Sony_IMX135_BLCCSC_isotropic.zip. __Citation__ Any work that uses the dataset should cite the following publication: F. Laakom, J. Raitoharju, J. Nikkanen, A. Iosifidis and M. Gabbouj, "INTEL-TAU: A Color Constancy Dataset," in IEEE Access, vol. 9, pp. 39560-39567, 2021
Näytä enemmän

Julkaisuvuosi

2019

Aineiston tyyppi

Tekijät

Aarhus University

Alexandros Iosifidis Orcid -palvelun logo - Muu tekijä

Tampereen yliopisto

Firas Laakom Orcid -palvelun logo - Julkaisija, Tekijä

Jenni Raitoharju Orcid -palvelun logo - Muu tekijä

Moncef Gabbouj - Muu tekijä

XIAOMI

Jarno Nikkanen - Muu tekijä

Projekti

Muut tiedot

Tieteenalat

Sähkö-, automaatio- ja tietoliikennetekniikka, elektroniikka

Kieli

Saatavuus

Avoin

Lisenssi

Creative Commons Nimeä JaaSamoin 4.0 Kansainvälinen (CC BY SA 4.0)

Avainsanat

Color constancy, computer vision, Illumination estimation

Asiasanat

konenäkö

Ajallinen kattavuus

undefined

Liittyvät aineistot