https://github.com/imshymike/handgestures
Python library that makes it easy to create and detect custom gestures using mediapipe
https://github.com/imshymike/handgestures
hand-gesture-recognition hand-tracking mediapipe mediapipe-hands opencv
Last synced: 5 months ago
JSON representation
Python library that makes it easy to create and detect custom gestures using mediapipe
- Host: GitHub
- URL: https://github.com/imshymike/handgestures
- Owner: ImShyMike
- License: mit
- Created: 2025-04-11T13:05:02.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2025-04-11T16:19:16.000Z (6 months ago)
- Last Synced: 2025-04-11T17:32:26.893Z (6 months ago)
- Topics: hand-gesture-recognition, hand-tracking, mediapipe, mediapipe-hands, opencv
- Language: Python
- Homepage:
- Size: 30.3 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# HandGestures
Python library that makes it easy to create and detect custom gestures using mediapipe.
## Example Image
(the image is using the 'floating_mouse.py' example with the debug option '_blank_frame=True')
## Example Code
```python
# Import the needed classes
from hand_gestures import Condition, GestureBuilder, GestureHandler
# First define a list of gestures
class Gestures:
NONE = "none"
FIST = "fist"
GRAB = "grab"
OK = "ok"
PINCH = "pinch"
TWO_UP = "two up"
# Then create the gesture handler (this will use the camera)
handler = GestureHandler(0)
# This function is an example for a callback
def callback_func(callback_dict):
print(
f"1 - Hand: {callback_dict[0]['hand']}, Distance: {callback_dict[0]['distance']}, Condition Median Point: {callback_dict[0]['condition_median_point']}"
)
# This callback will be used as the main callback and it is called on every frame
def main_callback_func(hand, detected_gestures, hand_scale, hand_positions):
# Using 'detected_gestures' you can set which gesture is being detected
# (this may look redundant but it is needed as sometimes multiple gestures are detected at once)
if not detected_gestures:
handler.set_gesture(hand, Gestures.NONE)
return
if Gestures.OK in detected_gestures:
# Here, FIST takes priority over OK (if both are present)
if Gestures.FIST in detected_gestures:
handler.set_gesture(hand, Gestures.FIST)
else:
handler.set_gesture(hand, Gestures.OK)
elif Gestures.GRAB in detected_gestures:
handler.set_gesture(hand, Gestures.GRAB)
elif Gestures.PINCH in detected_gestures:
handler.set_gesture(hand, Gestures.PINCH)
elif Gestures.TWO_UP in detected_gestures:
handler.set_gesture(hand, Gestures.TWO_UP)
# Then define all the gestures by using conditions
# (coordinates are in screen distance and are normalized from 0.0 to 1.0)
ok = GestureBuilder(Gestures.OK, [Condition(0, [8, 12, 16, 20], 0.15)])
# The structure is GestureBuilder(gesture name, [list of conditions])
# Each condition can have many settings but the required ones are:
# landmark1, landamark2, distance (landmark2 can be a list)
# A visual landmark cheatsheet can be found here:
# https://mediapipe.dev/images/mobile/hand_landmarks.png
fist = GestureBuilder(Gestures.FIST, [Condition(4, 6, 0.05)])
grab = GestureBuilder(
Gestures.GRAB, [Condition(4, 8, 0.05, hand="Left")], callback=callback_func
)
pinch = GestureBuilder(Gestures.PINCH, [Condition(4, 12, 0.05, hand="Right")])
two_up = GestureBuilder(
Gestures.TWO_UP,
[Condition(4, [16, 20], 0.1), Condition(8, 12, 0.05, invert_check=True)],
)
# Finally, start the running loop with all of your gestures (blocking)
handler.init([fist, ok, grab, pinch, two_up], main_callback_func)
```