perf: improved hsv color mapping in HeatMapAnnotator by 20x

[Armaggheddon]

Description

Replaced np.zeros() and subsequent fill operations with np.full(). This improves performance of the operation on a 1920x1080 frame by ~16x.

Type of change

• Performance improvement

How has this change been tested, please provide a testcase or example of how you tested the change?

Tested running in a loop using time.perf_counter() the default implementation with the updated one
Results are average over 10000 iterations

• with 3840x2160 frame: default takes average of 62ms, improved takes average of 5ms. Improvement of 11.5x
• with 1920x1080 frame: default takes average of 16ms, improved takes average 1ms. Improvement of 16x
• with 1280x720 frame: default takes average of 8ms, improved takes average 0.3ms. Improvement of 22.5x

All comparisons have been performed on the same hardware.

Any specific deployment considerations

No additional requirements

[CLAassistant]

All committers have signed the CLA.

[SkalskiP]

Hi @Armaggheddon 👋🏻 Thank you so much for your interest in supervision. It looks interesting. Could you share a script (preferably a Google Colab) where we could measure the performance boost?

[Armaggheddon]

Hi @SkalskiP !. I am happy to contribute😁. Sure below is the link to a colab notebook that implements just the core logic i committed and compares my vs the current implementation. The code should be pretty easy to follow. The notebook gives slightly different trends compared to the one i observed on my machine, however the improvement is still noticeable. Here is the LINK

These are the results I obtained running 100 iterations for each resolution:

Measuring perf with 
- (720, 1280, 3):
	- [BASE] Avg:  10.109914 ms, Total:  1010.991401 ms
	- [IMPROVED] Avg:  1.055095 ms, Total:  105.509520 ms
	==>  9.58x times faster
- (1080, 1920, 3):
	- [BASE] Avg:  98.214145 ms, Total:  9821.414546 ms
	- [IMPROVED] Avg:  3.421949 ms, Total:  342.194868 ms
	==>  28.70x times faster
- (2160, 3840, 3):
	- [BASE] Avg:  243.822512 ms, Total:  24382.251157 ms
	- [IMPROVED] Avg:  10.611352 ms, Total:  1061.135250 ms
	==>  22.98x times faster

Hope it helps!

[SkalskiP]

Thanks a lot @Armaggheddon! 🔥 Merging!