This package includes several deep learning models with demos of their use 
in Python and OpenCV, as well as the implementation of the CameraAgent and 
ActionAgent agents, from which you can derive ControlAgent or other agents. 
Pay attention to the achieved FPS, i.e., how many images per second you 
can process.

Projects (select only one):

• Implement mirroring of a person's emotion in front of a camera onto an iCub 
  robot in a simulator. Use face detection and emotion classification models 
  (Fig. schemes/emotion-mirroring.jpg).

• Implement that the robot counts on its fingers, imitating a person counting 
  on their fingers. Use the feature vector provided by the DINO model 
  (Fig. schemes/fingers-counting.jpg).

• Implement that the robot follows a ball moving on a table with its gaze. 
  Use the attention map provided by the DINO model 
  (Fig. schemes/following-by-gaze.jpg; the camera is an internal camera in the simulator).

• Implement the robot to seemingly track a nearby object shown to the camera 
  with the gaze. Use the attention map provided by the DINO model 
  (Fig. schemes/following-by-gaze.jpg; the camera is an external USB camera).

• Implement the imitation of a person in front of the camera by the simulated robot. 
  Use the feature vector provided by the DINO model and the encoder and decoder 
  of the robot's poses 
  (Fig. schemes/pose-imitation.jpg).