Deep Learning for Illuminant Estimation - Revision history

imported>Projects221: /* Appendix I */

2015-12-12T07:50:19Z

Appendix I

← Older revision		Revision as of 07:50, 12 December 2015
Line 265:		Line 265:

	[[File:SVM_BOF.zip]]		[[File:SVM_BOF.zip]]


			RGB images
			[[File:RGB.zip]]

			RAW images
			[[File:RAW.zip]]

	== Appendix II ==		== Appendix II ==

imported>Projects221: /* k-NN Results */

2015-12-12T07:32:05Z

k-NN Results

← Older revision		Revision as of 07:32, 12 December 2015
Line 140:		Line 140:

	Validation accuracies for RGB and RAW images using k-NN algorithm		Validation accuracies for RGB and RAW images using k-NN algorithm


	The plot shows validation accuracy at different values of k. Although accuracies up to k = 500 are computed, the most representative data are shown. The validation accuracy for RGB data peaks at k=15, and for RAW data peaks at k=30. The table summarizes training, validation and test accuracies of RGB and RAW data at the respective peak values of k. Since color histograms have larger size than the gray histograms, it is expected that the accuracies for RGB data is higher than those for RAW data. Better choice of k may be achieved with cross validation.		The plot shows validation accuracy at different values of k. Although accuracies up to k = 500 are computed, the most representative data are shown. The validation accuracy for RGB data peaks at k=15, and for RAW data peaks at k=30. The table summarizes training, validation and test accuracies of RGB and RAW data at the respective peak values of k. Since color histograms have larger size than the gray histograms, it is expected that the accuracies for RGB data is higher than those for RAW data. Better choice of k may be achieved with cross validation.

imported>Projects221: /* k-NN Results */

2015-12-12T07:29:56Z

k-NN Results

← Older revision		Revision as of 07:29, 12 December 2015
Line 141:		Line 141:
	Validation accuracies for RGB and RAW images using k-NN algorithm		Validation accuracies for RGB and RAW images using k-NN algorithm

	~~Figure 5~~ shows ~~valida- tion~~ accuracy at different values of k. Although accuracies up to k = 500 are computed, the most representative data are shown. The validation accuracy for RGB data peaks at k=15, and for RAW data peaks at k=30.~~Table2sum- marizes~~ training, validation and test accuracies of RGB and RAW data at the respective peak values of k. Since color histograms have larger size than the gray histograms, it is expected that the accuracies for RGB data is higher than those for RAW data. Better choice of k may be achieved with cross validation.		The plot shows validation accuracy at different values of k. Although accuracies up to k = 500 are computed, the most representative data are shown. The validation accuracy for RGB data peaks at k=15, and for RAW data peaks at k=30. The table summarizes training, validation and test accuracies of RGB and RAW data at the respective peak values of k. Since color histograms have larger size than the gray histograms, it is expected that the accuracies for RGB data is higher than those for RAW data. Better choice of k may be achieved with cross validation.

	The following table shows the training, validation and test accuracies of RGB and RAW data at the respective peak values of k.		The following table shows the training, validation and test accuracies of RGB and RAW data at the respective peak values of k.

imported>Projects221: /* k-NN Results */

2015-12-12T07:29:16Z

k-NN Results

← Older revision		Revision as of 07:29, 12 December 2015
Line 135:		Line 135:

	=== k-NN Results ===		=== k-NN Results ===
	In ~~our~~ k-NN ~~classifier~~, ~~we found that~~ the choice of k has a great impace on the performance, ~~and~~ choosing the optimal k ~~was~~ important to obtaining the best k-NN classifier. The below figure summarizes the validation accuracy of the classifier run on the test data at different values of k.		In k-NN classifiers, the choice of k has a great impace on the performance, hence choosing the optimal k is important to obtaining the best k-NN classifier. The below figure summarizes the validation accuracy of the classifier run on the test data at different values of k.

	[[File:Knnacc.png\|450px]]		[[File:Knnacc.png\|450px]]
Line 141:		Line 141:
	Validation accuracies for RGB and RAW images using k-NN algorithm		Validation accuracies for RGB and RAW images using k-NN algorithm

			Figure 5 shows valida- tion accuracy at different values of k. Although accuracies up to k = 500 are computed, the most representative data are shown. The validation accuracy for RGB data peaks at k=15, and for RAW data peaks at k=30.Table2sum- marizes training, validation and test accuracies of RGB and RAW data at the respective peak values of k. Since color histograms have larger size than the gray histograms, it is expected that the accuracies for RGB data is higher than those for RAW data. Better choice of k may be achieved with cross validation.
	~~From the plot~~, ~~we can observe that~~ the accuracy for RGB data ~~is a maximum~~ at k = 15, and ~~the accuracy value is 62.17%. The accuracy~~ for RAW data ~~is greatest~~ at k = 30, and the ~~value is 43.1%~~. ~~The higher accuracy for the RGB data is expected, since~~ color histograms have larger size than the gray histograms. ~~A better~~ choice ~~for~~ k may be achieved ~~through~~ cross validation.

	The following table shows the training, validation and test accuracies of RGB and RAW data at the respective peak values of k.		The following table shows the training, validation and test accuracies of RGB and RAW data at the respective peak values of k.

imported>Projects221: /* Acknowledgements */

2015-12-12T07:16:34Z

Acknowledgements

← Older revision		Revision as of 07:16, 12 December 2015
Line 234:		Line 234:
	== Acknowledgements ==		== Acknowledgements ==

	We would like to thank Brian Wandell, Joyce Farrell, and Rosemary Le for general support throughout the class and project, and special thanks to Henryk Blasinski for help with ~~image generation and account setup for Caffe~~.		We would like to thank Brian Wandell, Joyce Farrell, and Rosemary Le for general support throughout the class and project, and special thanks to Henryk Blasinski for help with the project.

	== References ==		== References ==

imported>Projects221: /* Data Generation */

2015-12-12T07:14:55Z

Data Generation

← Older revision		Revision as of 07:14, 12 December 2015
Line 59:		Line 59:
	[[File:camera_RAW.png\|550px]]		[[File:camera_RAW.png\|550px]]

	Input scenes under different illuminant after camera simulation of optics.		Input scenes under different illuminant after camera simulation of optics and sensor.

imported>Projects221: /* CNN Results */

2015-12-12T07:02:57Z

CNN Results

← Older revision		Revision as of 07:02, 12 December 2015
Line 193:		Line 193:
	The confusion matrices shown below were also obtained for the RGB and RAW images to give a better insight on the predictions of the classifier. The horizontal axes of the confusion matrices represent true labels of the illuminants, and the vertical axes the predicted illuminant labels. The numbers are normalized to the total number of images in the sets.		The confusion matrices shown below were also obtained for the RGB and RAW images to give a better insight on the predictions of the classifier. The horizontal axes of the confusion matrices represent true labels of the illuminants, and the vertical axes the predicted illuminant labels. The numbers are normalized to the total number of images in the sets.

	[[File:Cm256rgb4_test_644.png\|~~500px~~]]		[[File:Cm256rgb4_test_644.png\|460px]]

	Test Confusion Matrix for RGB data		Test Confusion Matrix for RGB data
Line 199:		Line 199:


	[[File:Cm256raw3_test_644.png\|~~500px~~]]		[[File:Cm256raw3_test_644.png\|460px]]

	Test Confusion Matrix for RAW data		Test Confusion Matrix for RAW data

imported>Projects221: /* Conclusions and Future Work */

2015-12-12T07:02:32Z

Conclusions and Future Work

← Older revision		Revision as of 07:02, 12 December 2015
Line 226:		Line 226:

	== Conclusions and Future Work ==		== Conclusions and Future Work ==
	In comparing the baseline k-NN classifier and the CNN classifier for illuminant estimation, the deep learning method of CNNs ~~achieve~~ higher accuracy. This is expected, because the network is able to learn different levels of features of an image. One major limitation during this project was the resource of time. As mentioned, the long training time hinders the optimal selection of hyperparameters to choose the most effective classifier model.		In comparing the baseline k-NN classifier and the CNN classifier for illuminant estimation, the deep learning method of CNNs achieved higher accuracy. This is expected, because the network is able to learn different levels of features of an image. One major limitation during this project was the resource of time. As mentioned, the long training time hinders the optimal selection of hyperparameters to choose the most effective classifier model.

	Besides trying different datasets to work with, future work for the CNN classifier should focus on ~~investigating~~ a better CNN architecture for classifying RAW images. ~~Other tasks~~ to experiment with would be ~~scrampbling~~ the ~~pixel~~ of input images to investigate the influence of spatial arrangement on the classifier performance.		Besides trying different datasets to work with, future work for the CNN classifier should focus on implementing a better CNN architecture for classifying RAW images. Another aspect to experiment with would be scrambling the pixels of input images to investigate the influence of spatial arrangement on the classifier performance.

	In terms of SVM algorithms, besides increasing the dataset size, further work can be done to explore what types of feature vectors are best used for this application. ~~We found that the~~ Bag of Features method did not work well with illuminant classification, since it is better at detecting textures and objects in the scene. The Histogram of Oriented Gradients feature set may also not work very well, since these mostly have ~~best~~ performance at finding edges of images. Since we are classifying illuminants, the range of colors would be a better feature space to focus on. Perhaps different color features like color histograms could be used along with the SVM to estimate the illuminant.		In terms of SVM algorithms, besides increasing the dataset size, further work can be done to explore what types of feature vectors are best used for this application. The Bag of Features method did not work well with illuminant classification, since it is better at detecting textures and objects in the scene. The Histogram of Oriented Gradients feature set may also not work very well, since these mostly have good performance at finding edges of images. Since we are classifying illuminants, the range of colors would be a better feature space to focus on. Perhaps different color features like color histograms could be used along with the SVM to estimate the illuminant.

	== Acknowledgements ==		== Acknowledgements ==

imported>Projects221: /* Conclusions and Future Work */

2015-12-12T07:00:56Z

Conclusions and Future Work

← Older revision		Revision as of 07:00, 12 December 2015
Line 230:		Line 230:
	Besides trying different datasets to work with, future work for the CNN classifier should focus on investigating a better CNN architecture for classifying RAW images. Other tasks to experiment with would be scrampbling the pixel of input images to investigate the influence of spatial arrangement on the classifier performance.		Besides trying different datasets to work with, future work for the CNN classifier should focus on investigating a better CNN architecture for classifying RAW images. Other tasks to experiment with would be scrampbling the pixel of input images to investigate the influence of spatial arrangement on the classifier performance.

	In terms of SVM algorithms, besides increasing the dataset size, further work can be done to explore what types of feature vectors are best used for this application. We found that the Bag of Features method did not work well with illuminant classification, since it is better at detecting textures and objects in the scene. The Histogram of Oriented Gradients may also not work very well, since these mostly have best performance at finding edges of images. Since we are classifying illuminants, the range of colors would be a better feature space to focus on. Perhaps different color features like color histograms could be used along with the SVM to estimate the illuminant.		In terms of SVM algorithms, besides increasing the dataset size, further work can be done to explore what types of feature vectors are best used for this application. We found that the Bag of Features method did not work well with illuminant classification, since it is better at detecting textures and objects in the scene. The Histogram of Oriented Gradients feature set may also not work very well, since these mostly have best performance at finding edges of images. Since we are classifying illuminants, the range of colors would be a better feature space to focus on. Perhaps different color features like color histograms could be used along with the SVM to estimate the illuminant.

	== Acknowledgements ==		== Acknowledgements ==

imported>Projects221: /* Conclusions and Future Work */

2015-12-12T07:00:43Z

Conclusions and Future Work

← Older revision		Revision as of 07:00, 12 December 2015
Line 230:		Line 230:
	Besides trying different datasets to work with, future work for the CNN classifier should focus on investigating a better CNN architecture for classifying RAW images. Other tasks to experiment with would be scrampbling the pixel of input images to investigate the influence of spatial arrangement on the classifier performance.		Besides trying different datasets to work with, future work for the CNN classifier should focus on investigating a better CNN architecture for classifying RAW images. Other tasks to experiment with would be scrampbling the pixel of input images to investigate the influence of spatial arrangement on the classifier performance.

	In terms of SVM algorithms, besides increasing the dataset size, further work can be done to explore what types of feature vectors are best used for this application. We found that the Bag of Features method did not work well with illuminant classification, since it is better at detecting textures and objects in the scene. ~~We also believe~~ Histogram of Oriented Gradients ~~would~~ not work very well, since these mostly have best performance at finding edges of images. Since we are classifying illuminants, the range of colors would be a better feature space to focus on. Perhaps different color features like color histograms could be used along with the SVM to estimate the illuminant.		In terms of SVM algorithms, besides increasing the dataset size, further work can be done to explore what types of feature vectors are best used for this application. We found that the Bag of Features method did not work well with illuminant classification, since it is better at detecting textures and objects in the scene. The Histogram of Oriented Gradients may also not work very well, since these mostly have best performance at finding edges of images. Since we are classifying illuminants, the range of colors would be a better feature space to focus on. Perhaps different color features like color histograms could be used along with the SVM to estimate the illuminant.

	== Acknowledgements ==		== Acknowledgements ==