Jeremy Janella
76 lines
2.5 KiB
Plaintext
76 lines
2.5 KiB
Plaintext
**********************************************
|
|
The German Traffic Sign Recognition Benchmark
|
|
**********************************************
|
|
|
|
This archive contains the training set of the
|
|
"German Traffic Sign Recognition Benchmark".
|
|
|
|
This training set is supposed be used for the online competition
|
|
as part of the IJCNN 2011 competition. It is a subset of the final
|
|
training set that will be published after the online competition is
|
|
closed.
|
|
|
|
|
|
**********************************************
|
|
Archive content
|
|
**********************************************
|
|
This archive contains the following structure:
|
|
|
|
There is one directory for each of the 43 classes (0000 - 00043).
|
|
Each directory contains the corresponding training images and one
|
|
text file with annotations, eg. GT-00000.csv.
|
|
|
|
|
|
**********************************************
|
|
Image format and naming
|
|
**********************************************
|
|
The images are PPM images (RGB color). Files are numbered in two parts:
|
|
|
|
XXXXX_YYYYY.ppm
|
|
|
|
The first part, XXXXX, represents the track number. All images of one class
|
|
with identical track numbers originate from one single physical traffic sign.
|
|
The second part, YYYYY, is a running number within the track. The temporal order
|
|
of the images is preserved.
|
|
|
|
|
|
**********************************************
|
|
Annotation format
|
|
**********************************************
|
|
|
|
The annotations are stored in CSV format (field separator
|
|
is ";" (semicolon) ). The annotations contain meta information
|
|
about the image and the class id.
|
|
|
|
|
|
In detail, the annotations provide the following fields:
|
|
|
|
Filename - Image file the following information applies to
|
|
Width, Height - Dimensions of the image
|
|
Roi.x1,Roi.y1,
|
|
Roi.x2,Roi.y2 - Location of the sign within the image
|
|
(Images contain a border around the actual sign
|
|
of 10 percent of the sign size, at least 5 pixel)
|
|
ClassId - The class of the traffic sign
|
|
|
|
|
|
**********************************************
|
|
Further information
|
|
**********************************************
|
|
For more information on the competition procedures and to obtain the test set,
|
|
please visit the competition website at
|
|
|
|
http://benchmark.ini.rub.de
|
|
|
|
If you have any questions, do not hesitate to contact us
|
|
|
|
tsr-benchmark@ini.rub.de
|
|
|
|
|
|
**********************************************
|
|
Institut für Neuroinformatik
|
|
Real-time computer vision research group
|
|
|
|
Ruhr-Universität Bochum
|
|
Germany
|
|
********************************************** |