light-sync: A Python script for Philips Hue

Sync Philips Hue lights with computer screen in real time

Prerequisites:

• Python environment
• Pip: apt get install python-pip
• OpenCV Python: pip install opencv-python
• MSS: pip install mss
• Requests: pip install requests

Usage:

Make sure you set the MY_LIGHT_NAMES/MY_LIGHT_IDS and BRIDGE_IP variables to match your Hue System.

Press the Hue Bridge button before running the script for the first time!

Run the script in the terminal:

python light_sync.py

For stereo mode start 2 instances of the script (make sure to divide by 2 the HUE_MAX_REQUESTS_PER_SECOND variable:

python light_sync.py --screenpart left --lights Light1
python light_sync.py --screenpart right --lights Light2

Feel free to tweak setup variables in order to obtain preferred effect.

Enjoy!

Desktop app for Linux, Mac, Windows available:

https://github.com/digital-concrete/light-sync-electron

How it works:

1. Use Python MSS to grab a screen capture

2. Shrink the captured image in order to increase computation speed. Check out INPUT_IMAGE_REDUCED_SIZE variable.

3. Use cv2.matchTemplate to check the difference between current and previously used frame. If the difference is smaller than the FRAME_MATCH_SENSITIVITY variable then the frame is skipped in order to prevent redundant calculations and requests. Let's suppose we have following frame:

4. Make a grayscale copy of the image and apply OpenCV threshold function in order to calcuate mask:

5. Apply mask to image:

6. Compare the count of non zero value pixels in the image with MIN_NON_ZERO_COUNT. If the count is too little, then turn off/dim the lights. If VARIABLE_BRIGHTNESS_MODE is set to true then brightness is calculated based on the non zero pixel count. If DIM_LIGHTS_INSTEAD_OF_TURN_OFF is set to true then, whenever the count is below the lower threshold MIN_NON_ZERO_COUNT, the lights are being dimmed instead of turned off.

7. Apply OpenCV K Means Clustering in order to find main image colors. Result will look like:

8. Calculate which of the colors calculated in step 3 should be sent to Philips Hue lights. If the most prevalent color is either too dark or too bright it means that we have an image with bright or dark background. In this case we look for the next color until we find a color that suits the conditions.

9. Check whether the calculated color is different than the one already in use. Tweak COLOR_SKIP_SENSITIVITY variable to change the sensitivity.

10. Finally, we have the CAN_UPDATE_HUE variable that allows us to update the lights. This variable is used to prevent bridge request bottlenecks and is cleared by a timeout. Timeout duration can be adjusted by changing HUE_MAX_REQUESTS_PER_SECOND variable.

11. If the above flag is clear we can finally update the lights: change color, brightness, dim or switch them on/off :)

Based on:

http://python-mss.readthedocs.io/examples.html

https://docs.opencv.org/3.0-beta/doc/py_tutorials/py_ml/py_kmeans/py_kmeans_opencv/py_kmeans_opencv.html

https://www.developers.meethue.com/

https://github.com/studioimaginaire/phue

https://github.com/benknight/hue-python-rgb-converter