GENUA PESTO
GENoa hUman Active fixation database: PEripersonal space STereoscopic images and grOund truth disparity

Reference publication (please cite it!)

A. Canessa, A. Gibaldi, M. Chessa, M. Fato, F. Solari, and S.P. Sabatini. (2017). A dataset of stereoscopic images and ground-truth disparity mimicking human fixations in peripersonal space. Scientific Data, 4. [doi]

Dataset download on Dryad

Dataset download

DISPARITY AND OPTIC FLOW - Public Evaluation Page

Please select where to compute indexes	Please select the image

ORIGINAL - IMAGE QUALITY INDEXES

Method	MAE			NCC			SSIM			Left Image	Ground truth	Estimate (horiz.)
Method	EST	GT	perc	EST	GT	perc	EST	GT	perc	Left Image	Ground truth	Estimate (horiz.)
SGMMAT	27.98	27.98	1.00	0.74	0.74	1.00	0.52	0.52	1.00
DISPOP	27.98	27.98	1.00	0.74	0.74	1.00	0.52	0.52	1.00
CLASSIC-NL_FAST	27.98	27.98	1.00	0.74	0.74	1.00	0.52	0.52	1.00
FARNEBACK	27.98	27.98	1.00	0.74	0.74	1.00	0.52	0.52	1.00

IMAGE QUALITY INDEXES
These indexes are used to describe the quality of the estimated disparity. The Right image is warped of an amount equal to the stereo disparity, and is then compared to the Left image. The computation is repated using the Ground-Truth disparity, for comparison.

MAE (Mean) Absolute Error between the pixel values of the original Left image and Right image, warped using the estimated stereo disparity.
MAE (GT) - between the pixel values of the original Left image and Right image, warped using the Ground-Truth stereo disparity.
MAE (perc) - percentage ration between the MAE and the MAE_GT
NCC (Mean)- Normalized Cross Correlation between the original Left image and Right image, warped using the estimated stereo disparity.
NCC (GT) - Normalized Cross Correlation between the original Left image and Right image, warped using the Ground-Truth stereo disparity.
NCC (perc) - percentage ration between the NCC and the NCC_GT
SSIM* - Structure SIMilarity index between the original Left image and Right image, warped using the estimated stereo disparity.
SSIM (GT) - Structure SIMilarity index between the original Left image and Right image, warped using the Ground-Truth stereo disparity.
SSIM (perc) - percentage ration between the SSIM_EST and the SSIM_GT

*H.R. Sheikh and A.C. Bovik. Image information and visual quality. IEEE Transactions on Image Processing, 15(2):430–444, 2006.

ORIGINAL - DISPARITY (OPTIC FLOW) INDEXES

Method	Horizontal est		Vertical est		Vector		NAN		Left Image	Ground truth	Estimate (horiz.)
Method	mean	std	mean	std	mean	std	X	Y	Left Image	Ground truth	Estimate (horiz.)

STEREO DISPARITY INDEXES

These indexes quantify the difference between the estimated disparity and the Ground-Truth disparity.

DX mean - mean of the error between the estimated and Ground-Truth HORIZONTAL dispairty
DX std - standard deviation error between the estimated and Ground-Truth HORIZONTAL dispairty
DY mean - mean of the error between the estimated and Ground-Truth VERTICAL dispairty
DV mean - mean of the error between the estimated and Ground-Truth VECTOR dispairty
DV std - standard deviation error between the estimated and Ground-Truth VECTOR dispairty

NAN_X - Percentage of invalid pixels in the horizontal disparity map
NAN_Y - Percentage of invalid pixels in the vertical disparity map

Algorithms in the table

SGMMAT - Semiglobal Matching - Current Implementation: Matlab R2016 Disparity class
DISPOP - Under Revision for International Journal of Neural Systems
CLASSIC-NL_FAST - Sun, D., Roth, S., & Black, M. J. (2014). A quantitative analysis of current practices in optical flow estimation and the principles behind them. International Journal of Computer Vision, 106(2), 115-137.
FARNEBACK - Farneback, G. "Two-Frame Motion Estimation Based on Polynomial Expansion." Proceedings of the 13th Scandinavian Conference on Image Analysis. Gothenburg, Sweden, 2003. Current Implementation: Matlab R2016 opticalFlowFarneback class

GENUA PESTO GENoa hUman Active fixation database: PEripersonal space STereoscopic images and grOund truth disparity