snavas / gecco Goto Github PK

Maybe to try a lightweight neural network (like YOLO?) with the colorimage, depthimage & fingertip features. Use current approach to generate training data.

On the new PC sometimes GECCO throws this error:

Process finished with exit code -1073741819 (0xC0000005)

Currently the 415 camera always throws this error while the 435 only spontaneously throws it.

Sometimes it helps to change the USB port of the camera.

Downgrading Vidgear to 0.1.8 did not fix the issue.

opencv provides a function called Canny to detect edges. maybe this could be used to improve the contours of the hand

The current solution is not optimal because it produces empty pixels in the color stream.
This thread describes a more elaborate solution but I dont understand it yet...

Currently the visualization is delayed and runs only at about 3 frames per second. It should be investigated if this can be improved.
One possible approach could be adjusting the networking.

On the new PC after a while this error is thrown when we want to read the colorframe or the depthframe from the camera:

[ERROR] "RuntimeError: Frame didn't arrive within 5000"

Ideas:

• saturation & brightness: black when far away, bright color when near

• transparency: more transparent the further away

• pro: doesnt occlude background

ideas:

• put aruco codes on hands
• put qr codes on hands to assign names etc. (not as robust as aruco?)

The convex hull indices are not monotonous, which can be in the case when the input contour contains self-intersections in function 'cv::convexityDefects'

In the meantime using cv2 4.2.0.34 is recommended

This bug/unexpected behaviour starts happening since cv2 version4.3.0.36

pip install opencv-contrib-python==4.2.0.34

AttributeError: module 'cv2.cv2' has no attribute 'compression_param'
Helper :: CRITICAL :: `compression_param` is not a valid OpenCV property!

AttributeError: module 'cv2.cv2' has no attribute 'compression_format'
Helper :: CRITICAL :: `compression_format` is not a valid OpenCV property!

vidgear version= 0.1.9
opencv version= 4.2.0.34

Now the system only logs the detected hands from their own device, but not the ones from the partner device, only video frames are shared and communication between partners. It would be ideal that video frames + hand/finger tracking data is also shared.

Workaround: Right now we have two separated log files (system1.log and system2.log) that should be merged into one file after each session before further analysis.