< Results

Estimated Illuminant Data

[Color Checker Set]  [Grey ball Set]  [SFU 321 Set]


                  all_image_names: {1xN cell}   : Base names of images
          groundtruth_illuminants: [Nx3 double] : Ground truth illuminant for each image
         estimated_illuminants_ML: [Nx3 double] : ML Estimate of illuminant with proposed method
                angular_errors_ML: [Nx1 double] : Corresponding Angular Errors

Only for Color Checker database:

   estimated_illuminants_GenPrior: [Nx3 double] : Illuminant estimate with prior over all images in training set
          angular_errors_GenPrior: [Nx1 double] : Corresponding Angular Errors
 estimated_illuminants_CategPrior: [Nx3 double] : Illuminant estimate with separate priors for indoor/outdoor
        angular_errors_CategPrior: [Nx1 double] : Corresponding Angular Errors
                         isindoor: [Nx1 logical]: Label per image, 0 - outdoor, 1 - indoor

Performance Tables: Color Checker Database

Collected by Gehler et al., we use the linear re-processed version from Shi and Funt. Results on other algorithms based on data from Gijsenij et al..

Error Quantiles

* Worst-25% error is the mean of the highest 25% error values.
Pair-wise comparison with Sign Test