[FEATURE] Clear up murky images for better visibility in dark rooms

[ivelin]

Is your feature request related to a problem? Please describe.
Sometimes cameras take dark or murky images due to poor lighting conditions or camera quality.
It makes it hard for users to see the objects detected in a timeline image view. Users also report that the detection rates drop significantly in dim lighting conditions.

Describe the solution you'd like
Apply image transformation that improves the visibility of objects to the human eye and test how/if it improves AI model detection rates.

Describe alternatives you've considered
The alternative is to use higher quality more expensive cameras or demand better lighting conditions in the installed environment.

Additional context
There are new ML models used in modern phones such as Pixel to improve image quality via software. Now there are also open source models available for this task:

https://github.com/sayakpaul/MIRNet-TFLite

https://github.com/dawnlh/low-light-image-enhancement-resources

https://github.com/aiff22/DPED

https://github.com/alexjc/neural-enhance

https://ai.googleblog.com/2018/03/semantic-image-segmentation-with.html

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.