Infrared-sequence

1.1 Description

The infrared data is collected in the motion capture system in the Chinese BeiDou industrial base, with the aim ofth e test of Thermal NeRF.

1.2 Sensor parameter

• Infrared Camera: optris PI 450i,382*288

• Rostopics of rosbag sequences: /optris/thermal_image

• Image format: 16UC1

1.3 Dataset sequences

1.3.1 2023-02-28

Sequence Name	Collection Date	Total Size	Duration	Temperature	Feature	Rosbag	GT
room_rot_01	2023-02-28	274.5MB	44.7s	-	shake	Rosbag	GT
room_rot_02	2023-02-28	589.9MB	96s	-	slow	Rosbag	GT
room_rot_03	2023-02-28	522.5MB	85s	-	shake, dynamic	Rosbag	GT
room_rot_04	2023-02-28	481.5 MB	78s	-	slow, static	Rosbag	GT
room_move_01	2023-02-28	725.3 MB	118s	-	by car	Rosbag	GT
room_move_02	2023-02-28	521.2MB	84s	-	shake, rapid	Rosbag	GT
room_move_03	2023-02-28	406.1MB	66s	-	shake, slow, slope	Rosbag	GT
room_move_04	2023-02-28	311.6MB	50.7s	-	shake, dynamic	Rosbag	GT
room_move_05	2023-02-28	447.9MB	72s	-	shake, dynamic	Rosbag	GT

1.3.2 2023-03-18

This time, we collect the data at four different temperature. Due to the power limit of the air condition and the openness of the site, the importance of temperature error should be taken into consideration. Also, each trajectory at different temperature can be different.

Sequence Name	Collection Date	Total Size	Duration	Temperature	Feature	Rosbag	GT
room_rot	2023-03-18	385.4MB	62s	17°C	shake	Rosbag	GT
room_move	2023-03-18	605.2MB	98s	17°C	shake	Rosbag	GT
room_rot	2023-03-18	466.6MB	76s	20°C	shake	Rosbag	GT
room_move	2023-03-18	788.8MB	128s	20°C	shake	Rosbag	GT
room_rot	2023-03-18	662.3MB	107s	25°C	shake	Rosbag	GT
room_move	2023-03-18	698.7MB	113s	25°C	shake	Rosbag	GT
room_rot	2023-03-18	296.9MB	48.3s	30°C	shake	Rosbag	GT
room_move	2023-03-18	494.6MB	80s	30°C	shake	Rosbag	GT

1.3.3 2023-04-07

As there was no significant impact on the pixel distribution of the infrared photo from adjusting the temperature of the air conditioner last time, we are considering reconstructing an object with a significant temperature difference. We will fill a cup with water at different temperatures to adjust the temperature of the cup, and then take 360-degree multi-angle shots around the cup. It is worth noting that during the shooting process, the reflection point fell off the camera, so we reconstructed the rigid body multiple times, and the data from the same rigid body will be labeled.

Sequence Name	Collection Date	Total Size	Duration	Temperature	Feature	Rosbag	GT
water_scene_360_1	2023-04-07	1.3GB	217s	Hot	Rigid_body_1, multi-view	Rosbag	GT
water_scene_360_2	2023-04-07	1.3GB	213s	Warm	Rigid_body_1, multi-view	Rosbag	GT
water_scene_360_cp_1	2023-04-07	437.0MB	71s	30°C	Rigid_body_1, rotate	Rosbag	GT
water_scene_360_cp_2	2023-04-07	493.9MB	80s	30°C	Rigid_body_1, rotate	Rosbag	GT
scene_360_1	2023-04-07	2.0GMB	328s	30°C	Rigid_body_2, multi-view	Rosbag	GT
scene_360_2	2023-04-07	753.7MB	122s	30°C	Rigid_body_2, multi-view	Rosbag	GT
scene_360_3	2023-04-07	1.3GB	220s	20°C	Rigid_body_3, rotate	Rosbag	GT
scene_360_4	2023-04-07	755.6MB	123s	20°C	Rigid_body_3, multi-view	Rosbag	GT

1.3.4 2023-04-26

In this data collection, we once again shot a cup containing three different temperature water by rotating 360 degrees around it. However, it's worth noting that this time, when creating the rigid body, the position of the rigid body coordinate system and the camera coordinate system were rotated 180 degrees by placing the camera lens facing the back of the motion capture room screen. This made it more convenient for us to handle the conversion between the camera coordinate system and the rigid body coordinate system, and to avoid situations where markers fell during the shooting process and required the rigid body to be re-established.

Sequence Name	Collection Date	Total Size	Duration	Temperature	Feature	Rosbag	GT
water_scene_360_1	2023-04-26	1.7GB	283s	Very hot	Rigid_body_1, two views	Rosbag	GT
water_scene_360_2	2023-04-26	2.2GB	368s	Hot	Rigid_body_1, two views	Rosbag	GT
water_scene_360_3	2023-04-26	1.9GB	312s	Warm	Rigid_body_1, two views	Rosbag	GT

1.3.5 2023-05-13

This time, we ignore the temperature factor and only focus on the object and scene reconstruction. We adopt three different objects: two cups (one water bottle and one mug), a router, and a person. Also, according to the requirements of the LLFF data format, we shot a scene containing several warm parts by moving the camera within the same plane.

We collected the data in room 309, and the palette scaling minimal and maximal temperature of the camera launch file is set to $10$ and $60$ respectively.

Sequence Name	Collection Date	Total Size	Duration	Feature	Rosbag	GT
two_cups	2023-05-13	1.2GB	202s	Rigid_body, 360 deg, multiple views	Rosbag	GT
router	2023-05-13	623.1MB	101s	Rigid_body, < 180deg, multiple views	Rosbag	GT
human	2023-05-13	766，5MB	124s	Rigid_body, 180 deg, multiple views	Rosbag	GT
LLFF_scene	2023-05-13	475.9MB	77s	Rigid_body, within a plane	Rosbag	GT

1.3.5 2023-05-30

This time, we collected 5 sets of data at the Beidou Industrial Base. In the first two sets of data, we positioned the visible light camera and the infrared camera facing the same direction and maintained their relative positions unchanged. We conducted 360-degree rotational filming around three cups. In the remaining three sets of data, we used the infrared camera to capture data in two different formats: LLFF and nerfpp, for both the cups and people.

Sequence Name	Collection Date	Total Size	Duration	Feature	Rosbag	GT
three_cups_1	2023-05-30	7.30GB	262s	RGB, IR, 360 deg, single view, single height	Rosbag	GT
three_cups_2	2023-05-30	4.09GB	212s	RGB, IR, 360 deg, multi views, single height	Rosbag	GT
three_cups_3	2023-05-30	812MB	123s	IR, 360 deg, multi views, multi heights	Rosbag	GT
three_cups_4	2023-05-30	752MB	87s	IR, LLFF, within a plane	Rosbag	GT
person	2023-05-30	572MB	56s	IR, LLFF, within a plane	Rosbag	GT