RTC Dataset

Air Writing is a new writing technique where the character is written in the air and the character is recognized by the computer or mobile phone. Air writing is different from the traditional writing style.

Writing in the air can be defined as to write digit or character in a 3D space by using a finger or marker movement. Air-writing is different from the traditional writing style.  Due to the extensive improvement of sensor technologies, it has become prominent to track finger and joint precisely.

We have collected 30k trajectory data from various people. An Intel RealSense SR300 camera has been used to collect the data. We have named this RealSense based 3D trajectory character dataset as RTC.

Intel-Realsense-camera-coordinate

Intel-Realsense-camera-coordinate

The coordinate system for the RealSense camera is shown above. Z-axis is towards the observer.

The writing order is similart to unistroke style-

Download the dataset: Google Drive, Dropbox

Understanding datasets:

The dataset contains two files. One is for the feature and the other one is for the label.

label: The label is one hot vector.

Features: Contains the original data as a sequence of trajectory. Each tuple represents one feature. The sequence is stored as : x_0,y_0,z_0,x_1,y_1,z_1,......x_n,y_n,z_n

Here n is the length of that sequence.

How to use this dataset:

The data has been published in binary form. You have to pickle it first. If you don’t know how to pickle, no problem, here is the sample code –

import numpy as np 
import pickle 

#opening features 
with open(‘features’, ‘rb’) as fp: 
features=pickle.load(fp) 

#opening labels 
with open(‘labels’, ‘rb’) as fp: 
labels=pickle.load(fp)

Thank you for showing interest in this research.