GAP-LP dataset is released for academic research only, and is free to researchers from educational or research institutes for non-commercial purposes. In no event shall the Digital research center of Sfax (CRNS) be liable to any party for direct, indirect, special, incidental, or consequential damages arising out of the use of the data. CRNS specifically disclaims any warranties. The data provided hereunder is on an "as is" basis and the CRNS has no obligation to provide maintenance, support, updates, enhancements, or modifications.
The GAP-LP dataset images were acquired with different quality cameras under different resolutions, view angles and daylight lighting conditions to enable robust LP detection/recognition algorithms testing. GAP-LP dataset is freely available to research community. It is composed of 9,175 fully annotated images for both LP detection and recognition. The dataset is split as follows : 7117 images for training, 456 for validation and 1602 images for test.
LP detection
LP recognition
Every image is annotated with a text file containing the LP position using the following format:
0 x y h w
0 : index of the object
x = (((xmax + xmin) / 2) - 1) / width
y = (((ymax + ymin) / 2) - 1) / height
w = (xmax - xmin) / width
h = (ymax - ymin) / height
(xmin,ymin) is the left top corner, (xmax,ymax) is the right bottom corner and width and height are respectively the width and height of the image.
Every LP image is annotated with a text file containing their respective components bounding box coordinates. Every text file contains multiple lines (number of components in the license plate) in the following format:
0 x y h w
0 : index of the object(0,1,2,3,4,5,6,7,8,9,10...)
x = (((xmax + xmin) / 2) - 1) / width
y = (((ymax + ymin) / 2) - 1) / height
w = (xmax - xmin) / width
h = (ymax - ymin) / height
(xmin,ymin) is the left top corner, (xmax,ymax) is the right bottom corner and width and height are respectively the width and height of the image.