Fast Access

Expert Imitates NAO Dataset (EXIN)

Expert Imitates NAO Dataset (EXIN)

NAO-Based Imitation-Based Data Collection (Subject-Based Method)

In this method, a purely observational and imitation-based data collection process was employed, rather than relying on direct mathematical transformations, for training an artificial neural network (ANN) model. Specifically, 8 predefined movements of the NAO robot were recorded using the Choregraphe software, and these movements were then imitated one-to-one by human subjects.

While the subjects synchronized their actions to imitate NAO's movements, each motion sequence was recorded frame-by-frame both from the NAO and the human participants. As a result, a total of 600 data samples were collected. During the training phase, 80% of the data (480 samples) were used for training, while 20% (120 samples) were reserved for testing.

The accuracy of the model was evaluated by comparing the predicted test sequences against NAO’s original motion recordings. Metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE) were used to quantify the similarity between the predicted and actual movements.

The advantage of this dataset is that it directly contains human-performed motions, independent of teleoperation or mathematical transformation techniques. However, human-specific variations and errors during imitation may be directly reflected in the model.


What the file names in the dataset represent:

For example, let us consider the file named:
N01_C011_R01_fixed_left.npy

  • N01_: This part represents the subject code. The prefix “N” is used to indicate that the movement is based on imitation of NAO.

  • C011_: Indicates the code of the NAO movement that was imitated. For example, C011 refers to the "Confused" gesture, which is one of NAO's built-in motions from the Choregraphe software.

  • R01_: Refers to the repetition number of the performed action.

  • fixed_: Indicates that the data has been converted from JSON to NumPy format.

  • left: Specifies that the file contains data from the left arm.


Motion codes in the labeled dataset and their corresponding gestures:

C011_ = CONFUSED
C012_= ELLER YERE PARALEL
C013_= KISSES
C014_= LEFT HAND PARALEL
C015_= MYSTİCAL
C016_= TAKE PİCTURE
C017_= TURN RİGHT
C018_= PAKET KALDIRMA

 

Source Codes and Datasets