Updating thermal imaging dataset of hand gestures with unique labels

An update to the previously published low resolution thermal imaging dataset is presented in this paper. The new dataset contains high resolution thermal images corresponding to various hand gestures captured using the FLIR Lepton 3.5 thermal camera and Purethermal 2 breakout board. The resolution of the camera is 160×120 with calibrated array of 19,200 pixels. The images captured by the thermal camera are light-independent. The dataset consists of 14,400 images with equal share from color and gray scale. The dataset consists of 10 different hand gestures. Each gesture has a total of 24 images from a single person with a total of 30 persons for the whole dataset. The dataset also contains the images captured under different orientations of the hand under different lighting conditions.


Specifications Table
This section list the details of the hardware, procedure for collecting the data, and the format of the data.

Value of the Data
• The existing dataset contains the images of 32 × 32 pixel thermal camera [1,2] . However, the new dataset is created with 160 × 120 pixel thermal camera. • Efficient machine learning models can be developed to process the data for hand gesture recognition. • The academic or research communities working on thermal imaging data with efficient machine learning algorithms for hand gesture recognition or classification. • The data is also helpful in developing and testing efficient algorithms to work on thermal imaging dataset. • The data is collected with high thermal camera with no constraints on the environment and captured images are independent of back ground lighting conditions. This will be helpful for testing the algorithms with thermal imaging data.

Data Description
The previous dataset in [1,2] is captured with a low resolution thermal camera of 32 × 32 pixels resolution. The thermal images in this dataset correspond to ten hand gestures representing 0 to 9 sign language digits. This dataset has been created from various people with different hand orientations. On the other hand, the dataset present in this paper contains the images captured from the thermal camera with the resolution of 160 × 120 pixels. The thermal images contain ten different hand gestures captured from various people. We also captured images of both color and gray scale under varying environment conditions. Further, different hand orientations are also considered while creating the dataset.   Fig. 2 c, 2 d, 2 e, 2 f, 2 g, 2 h, 2 i, and 2 j are the color scale thermal images c, d, e, f, g, h, i, and j, respectively. Fig. 3 shows the gray thermal images corresponding to hand gestures from a to j. Fig. 3 a is corresponding to image a and Fig. 3 b corresponds to image b. Fig. 3 c, 3 d, 3 e, 3 f, 3 g, 3 h, 3 i, and 3 j are the gray scale thermal images c, d, e, f, g, h, i, and j, respectively.   Fig. 4 shows the experimental setup of the thermal camera considered for the data collection. We used FLIR Leptop 3.5 thermal camera module to be fitted in embedded systems [4] . It has a horizontal field of view of 57 • which indicates that it captures more of image details than object details. The images can be captured by connecting the thermal camera to a computer with Windows OS and SDK [5] . A portable and simple setup is to connect the thermal camera to Raspberry Pi 4 Model B [6] which make use of the python script to capture the images. For the creation of the dataset, we connected the thermal camera to Raspberry Pi 4 model B.

Experimental Design, Materials and Methods
The thermal camera do not have a port to directly connect to the Raspberry Pi. Thus, we fitted the camera on purethermal 2 breakout board [7,8] which has an interface to connect to the Raspberry Pi through USB. Thereafter, we placed the breakout board on a fixed stand to stable the camera as well as to add height for a easy capturing of the hand gestures. Fig. 5 shows the steps in the python script to capture the images for the dataset. The program makes use of the Lepton library from flirpy in Python. This library enables the thermal camera for capturing the images and then the OpenCV and matplotlib libraries are used to save the images in the Raspberry Pi. The main loop takes the inputs from 1 to 5 to capture the images. When 2 is the input, it will take second input. The second input defines how many images the thermal camera should capture before asking for the first input again.

Ethics Statement
The data consists solely of hand gestures and contains no personal information. It was a freefor-all campaign, and people gave the hand gestures at their own discretion.

Declaration of Competing Interest
The authors declare that there is no influence from known competing financial interests or personal relationships which have, or could be perceived for the work reported in this article.