ICDAR2023 Competition on Detection & Classification of Writing in Air

DataSet folder contains two subfolders DataSet_1 and DataSet_2 each for the specific task mentioned in the dataset.

TASK-1: CASSIFICATION OF WRITING IN AIR Aim of this task is to map air-written trajectories onto respective digits(0-9) and characters (A-Z and a-z).

Data provided for this task is presented in Dataset_1 folders and details of data are given below:

       1) It contains the data for training as Train_Set divided into Two sub folders (X,Y)_Co-ordinates and Sensor data.
             1a). (X,Y)_Co-ordinates : contains the x,y co-ordinates of reconstructed trajectories. 70 participants data is stored in separated folder with its ID and labels for it trajectory is mentioned in file name as well.

         1b). Sensor : 
                      Contains the data from same number of participants. The data is directly collected from IMU sensor and is each sensor data file is presented with its trajectory recontruction.

TASK-2 Aim of this task is to segment the writing from other gestures performed in air as binary classification task.

Data provided for this task is presented in Dataset_2 folders and details of data are given below:

       1) It contains the data for training as Train_Set divided into Two sub folders (X,Y)_Co-ordinates and Sensor data.
             1a). (X,Y)_Co-ordinates : 
                      Contains 38 participants and each participants data is stored in separated folder id and that folder contains the (X,Y)Co-Ordinates calculated or extracted using raw sensor data in Not_writing_sequence_start - writing_sequence - Not_writing_sequence_end 

         1b). Sensor : 
                      Contains 38 participants and each participants data is stored in separated folder id and that folder contains the selected raw feature data taken from raw sensor data in Not_writing_sequence_start - writing_sequence - Not_writing_sequence_end along with respective image as well.

Evaluation Metrics: Accuracy, Precision, and Recall

WE ARE LOOKING FORWARD TO....

(1). An executable program of complete pipeline from data preprocessing to result generation, requiring only path to input data (source files) and to store the output results (destination)
(2). One page explaination of your system implementation and methodology.

For Queries , please write us an email on junaid.younas@dfki.de