Outdoor Dataset for Place Categorization

Multi-Modal


Description

This dataset consists of 650 static panoramic scans containing high resolution dense 3D point clouds that are obtained using a FARO Focus 3D laser scanner and a Kodak PIXPRO SP360 camera mounted on a car. Each panoramic scan is composed of three synchronized sensor modalities: depth, reflectance, and RGB color. The FARO Focus3D sensor has a maximum range of 150 meters and a field of view of 360 degrees horizontally, and 300 degrees vertically. Each panoramic scan has a resolution of 5140×1757 pixels with 0.07 degrees horizontal and 0.17 degrees vertical angular resolution.

For each scan we moved the car 10-100 meters, stopped and took the full panoramic scan and color images. We repeated this process several times in each area. Point clouds were synchronized with color images off-line by using the SCENE software provided by FARO.



Data format

Each panoramic image is represented using the following PTX image format. PTX is an ASCII based interchange format for point cloud data. PTX file includes headers and point data with 7 columns. For a more complete description of the PTX format, please check here.

A PTX point cloud file contains a header and a body. The body starts in a new line:

		header
		body
		

PTX point cloud header format. Each row starts in a new line:

		number of columns 
		number of rows 
		st1 st2 st3 ; scanner registered position
		sx1 sx2 sx3 ; scanner registered axis 'X'
		sy1 sy2 sy3 ; scanner registered axis 'Y'
		sz1 sz2 sz3 ; scanner registered axis 'Z'
		r11 r12 r13 0 ; transformation matrix
		r21 r22 r23 0 ; this is a simple rotation and translation 4x4 matrix
		r31 r32 r33 0 ; just apply to each point to get the transformed coordinate
		tr1 tr2 tr3 1 ; use double-precision variables
		

PTX point cloud body format. Each row starts in a new line:

		X1 Y1 Z1 intensity R1 G1 B1   ; 3D coordinates intensity and RGB color  
		X2 Y2 Z2 intensity R2 G2 B2   ; Intensity in range: 0 - 1
		... 
		

Point Cloud Library (PCL) format

PTX files can be converted into PCL format using the following provided software:
ptxrgb2pcd.zip



Place Categories

The 3D color panoramic scans were obtained at six different place categories in Fukuoka city, Japan. The six categories include indoor parking, outdoor parking, coast, forest, residential area, and urban area, as shown in the map.



Each category contains 7 sets of panoramic images. Each set was obtained in a different place inside the same category. The following table summarizes the data:


Number of panoramic scans
Place Set 1 Set 2 Set 3 Set 4 Set 5 Set 6 Set 7 Total
Indoor Parking 16 16 13 15 17 13 15 105
Outdoor Parking 15 17 16 15 15 14 16 108
Coast 14 14 16 12 17 14 16 103
Forest 16 16 17 18 16 16 17 116
Residential 14 16 14 15 16 15 16 106
Urban 16 17 16 16 15 16 16 112
Total 91 96 92 91 96 88 96 650



Dataset

The complete dataset is available under request to dbadmin@irvs.ait.kyushu-u.ac.jp.

Data are stored using the following directory and files structure, where names in [brackets] indicate variables:

	
		[Category]  # Directory 
			|
			----- [Category].[PlaceId]_Scan_[ScanId].partial  # Directory containing partial color images
			|
			|
			----- [Category].[PlaceId]_Scan_[ScanId].Dep.png  # Depth Panoramic image. 
			|                                                 # Lighter colors indicate closer points.
			|
			----- [Category].[PlaceId]_Scan_[ScanId].ptx      # PTX file
			|
			----- [Category].[PlaceId]_Scan_[ScanId].Ref.png  # Reflectance image. 
 			|                                                 # Lighter values indicate higher reflectance.
			|
			----- [Category].[PlaceId]_Scan_[ScanId].RGB.png  # Color Panoramic image.
			|
			|
			----- [Category].[PlaceId]_Scan_[ScanId]_low.ptx  # Sampled point cloud (1284x440).
	

The following table shows the distribution of the dataset. An example is provided for each place category. Please, contact dbadmin@irvs.ait.kyushu-u.ac.jp to receive instructions to download the full dataset.

Dense Multi-modal Panoramic 3D Outdoor Dataset
Category
RGB Example Image (click for high resolution image) Depth Example Image (click for high resolution image) Reflectance Example Image (click for high resolution image) 3D movie (link to Youtube video)
Indoor parking
Example file: ParkingIn.zip (Example: Size 227 Mbytes)
Data size: 38.2 Gbyte, 105 scans
Map of places
Outdoor parking
Example file: ParkingOut.zip (Example: Size 148 Mbytes)
Data size: 40.0 Gbyte, 108 scans
Map of places
Coast area
Example file: Coast.zip (Example: Size 137 Mbytes)
Data size: 38.7 Gbyte, 103 scans
Map of places
Forest area
Example file: Forest.zip (Example: Size 176 Mbytes)
Data size: 43.2 Gbyte, 116 scans
Map of places
Residential area
Example file: Residential.zip (Example: Size 142 Mbytes)
Data size: 39.6 Gbyte, 106 scans
Map of places
Urban area
Example file: Urban.zip (Size 110 Mbytes)
Data size: 42.3 Gbyte, 112 scans
Map of places