adore-hrzz/nao-imitation

nao-imitation's Issues

Consolidate use of direction/hand

Not unsubscribing from the video device

Unsubscribe fails when exiting the program prematurely

Remove hard-coded parameters

There are several lines of code where paths and object names are hard coded. These should be removed and moved to the config files.

Remove object searching

In the current implementation, robot searches for the object by turning the head. This should be removed as the object will always be in front of the robot.

Load behaviors directly from .gst file

Currrently, each behavior is loaded individually, in imitation.py (~line 55, marked with TODO). This defeats the purpose of a configuration file, because changes to the configuration file are not reflected directly in the program.

OpenCV 3 compatibility

Image processing crashes when using OpenCV 3

State machine fails to exit

Once the whole procedure is performed, state machine fails to exit.

Integrate gesture recognition into the state machine

Remove external call to the separate gesture recognition script.

Remove writing of images to the disk

There are multiple imwrite calls in the code, most of which are used for debugging purposes and may not be entirely necessary.

Consolidate use of the ObjectTrackerModule class and nao-object-gesture module

Throughout the project, instance of the ObjectTrackerModule class is used to connect to the nao-object-gesture module running on the robot, which seems inefficient due to multiple loads and unloads and increased network traffic.
Issue will be updated with exact steps to fix it as they emerge.

Simple GUI for controlling the state machine

At least look into this. Might be too much work...

user_quit from the init state does not go to end state

@dmiklic can you check how does the state machine exit when you produce user_quit in the initial state? It does not seem to go through end state.
The importance of going through end state is that we need cleanup code to run there in order to fix #8.

Grab point identification

There are issues in the grab point identification algorithm, cause of which still needs to be identified.
One of the issues is related to the invalid values (NaNs in lists of doubles):
/..../NaoImageProcessing.py:110: RuntimeWarning: invalid value encountered in double_scalars

Clean up image processing code

• remove obsolete code (histogram based thresholding etc.)
• clean up debug output
• change integer constants into descriptive strings in config file

Update Readme

Segmentation sometimes fails

Ubrzati sve

Pauze izmedju pojedinih akcija su i dalje preduge