FC^4: Fully Convolutional Color Constancy with Confidence-Weighted Pooling

Single image portrait relighting

“…The final encoded activations at the bottleneck of the network are then used to predict the illumination corresponding to the source imageL s . This illumination prediction is performed using the "confidence learning" approach [Hu et al 2017], which has previously been used for the related problem of color constancy. This approach is used because, due to the finite spatial support of the conv layers in our network, the encoded network activations only reflect a limited region of the whole image -they can only "see" what lies within their receptive field.…”

Color Constancy Convolutional Autoencoder

“…C3AE fine-tuned and C3AE composite-loss achieve similar results, with C3AE fine-tuned performing better in terms of the mean error metric and C3AE composite-loss performing better in the mean of the worst 25%. [1] 154k Fc4(SqueezeNet) [2] 1.9M FC4 (AlexNet) [2] 3.8M DS-Net [3] 17.3M C3AE, fine-tuned 146k C3AE, composite-loss 146k…”

“…Barron [4] reformulated the problem of color constancy as a 2D spatial localization task, in order to directly learn how to discriminate between correctly white-balanced images and poorly whitebalanced images. Another CNN-based approach was proposed by Hu et al [2]. They introduced a novel pooling layer, namely Confidence-weighted pooling layer in an end-to-end learning process.…”

“…For the first experiment, we used ColorChecker RECommended dataset. Similarly to [1], [2], we used a three-fold cross validation on the folds provided with the dataset: for each run, one is used for training, one for validation, and the remaining one for testing For the second experiment, we used only Canon field2 set for training and validation (80% for training and 20% for validation). We constructed two test sets.…”

Flash Lightens Gray Pixe

“…3. Qualitative comparison on estimated illumination of flash gray pixels, with N light sources (2,3,5,8).…”