Pedestrian Lane Detection and Vanishing Point Estimation Dataset
1. Introduction
The pedestrian lane detection and vanishing point estimation (PLVP) dataset is created to support research on finding the pedestrian lane and estimating the vanishing point in an image.
It contains 2,000 color images that are captured in diverse environmental conditions: indoor, outdoor, different times of days, different weather, different background scenes. The dataset also includes the detection ground-truth for pedestrian lane and image vanishing point.
An example image from the dataset is shown in Fig. 1.
The dataset was prepared by M. C. Le, S. L. Phung, A. Nguyen, X. Wei, and X. Tang in 2012-2016.
(a) original image
|
(b) segmented pedestrian lane
|
(c) image vanishing point (red circle)
|
Figure 1: An example from the PLVP dataset. |
2. Dataset Description
The PLVP dataset (1.08 GB) has three sets, as shown in see Table 1.
Table 1: Format of the PLVP dataset.
Set |
Description |
File Name |
File Format |
1 |
Original color images |
xxxxxx.jpg |
JPEG color image |
2 |
Lane segmented images |
xxxxxx_lane.png |
PNG binary image |
3 |
Vanishing point files |
xxxxxx_vp.txt |
Text file |
- Set 1 consists of 2,000 original color images. The image files are in JPG format, and are named as xxxxxx.jpg, where xxxxxx is a six-digit number between 000001 and 002000.
- Set 2 contains the lane detection ground-truth, which is stored as PNG binary image files. In a PNG binary image file, a value of 1 indicates a pedestrian lane pixel whereas a value of 0 indicates a background pixel.
- Set 3 contains the vanishing point estimation ground-truth, which is stored as text files.
Each text file has the x and y coordinates of the image vanishing point.
For example, the vanishing point file 000227.txt for the image shown in Fig. 1(a) is as follows
That is, the vanishing point for this image is at x = 784.0 and y = 302.0.
- The example MATLAB code to read the dataset is included in subfolder 'PLVP-code'.
3. Download
For download links, please complete the Release Agreement form and email to phung at uow.edu.au.
4. Code to read the dataset
The MATLAB code to read the dataset is included in subfolder 'Code'.
- main_example.m Main example
- get_lane_groundtruth.m Function to read lane ground truth
- get_vp_groundtruth.m Function to read vanishing point ground truth
- apply_bin_white.m Miscellaneous function
|