ISETHDR CV Experiment - Revision history

Graykim: /* Appendix */

2025-12-09T20:29:04Z

Appendix

← Older revision		Revision as of 20:29, 9 December 2025
Line 420:		Line 420:

	== Appendix ==		== Appendix ==

	~~[[File:Snip 20210106183207.png\|200px]]~~

Graykim: /* YOLO */

2025-12-09T20:26:44Z

YOLO

← Older revision		Revision as of 20:26, 9 December 2025
Line 122:		Line 122:
	* Confidence interval		* Confidence interval
	Heatmaps are generated to visualize how confidence varies with exposure time across sensors.		Heatmaps are generated to visualize how confidence varies with exposure time across sensors.

			Below is the script used to iterate over the sensors and running YOLO on the ISET generated PNGs.

	<pre><nowiki>		<pre><nowiki>
Line 375:		Line 377:
	main()		main()
	</nowiki></pre>		</nowiki></pre>


	[[File:YoloSigmoid.png\|200px]]		[[File:YoloSigmoid.png\|200px]]

Graykim: /* YOLO */

2025-12-09T20:25:17Z

YOLO

Show changes

Graykim: /* Autonomous Vehicle Imaging Systems */

2025-12-09T20:12:11Z

Autonomous Vehicle Imaging Systems

← Older revision		Revision as of 20:12, 9 December 2025
Line 14:		Line 14:
	* Navigation under Autopilot		* Navigation under Autopilot
	A typical AV uses 8–9 cameras to provide a complete 360° view, making reliable image processing essential.		A typical AV uses 8–9 cameras to provide a complete 360° view, making reliable image processing essential.

	[[File:CameraSelfDrive.png]]		[[File:CameraSelfDrive.png]]

Graykim: /* HDR - High Dynamic Range */

2025-12-09T20:12:00Z

HDR - High Dynamic Range

← Older revision		Revision as of 20:12, 9 December 2025
Line 28:		Line 28:
	* SDR collapses detail near shadows and highlights → degrading YOLO detection.		* SDR collapses detail near shadows and highlights → degrading YOLO detection.
	Our challenge: Represent and manipulate HDR conditions using ISETCam tools.		Our challenge: Represent and manipulate HDR conditions using ISETCam tools.

	[[File:HDR.png]]		[[File:HDR.png]]

Graykim: /* YOLO */

2025-12-09T20:11:47Z

YOLO

← Older revision		Revision as of 20:11, 9 December 2025
Line 120:		Line 120:
	* Confidence interval		* Confidence interval
	Heatmaps are generated to visualize how confidence varies with exposure time across sensors.		Heatmaps are generated to visualize how confidence varies with exposure time across sensors.


	[[File:YoloSigmoid.png\|200px]]		[[File:YoloSigmoid.png\|200px]]

Graykim: /* Conclusions */

2025-12-09T20:11:37Z

Conclusions

← Older revision		Revision as of 20:11, 9 December 2025
Line 154:		Line 154:
	* Automotive HDR sensors consistently outperform consumer smartphone sensors for AV tasks.		* Automotive HDR sensors consistently outperform consumer smartphone sensors for AV tasks.
	* Split-pixel architectures (e.g., OV2312) provide the most robust overall performance.		* Split-pixel architectures (e.g., OV2312) provide the most robust overall performance.

	[[File:ResultsConfidence.png\|500px]]		[[File:ResultsConfidence.png\|500px]]

	[[File:CombinedConfidenceIntervals.png\|650px]]		[[File:CombinedConfidenceIntervals.png\|650px]]

	[[File:SensorDiscussion.png\|500px]]		[[File:SensorDiscussion.png\|500px]]

Graykim: /* Introduction */

2025-12-09T20:07:31Z

Introduction

← Older revision		Revision as of 20:07, 9 December 2025
Line 1:		Line 1:
	== Introduction ==		== Introduction ==
	Autonomous vehicles rely heavily on robust perception systems to interpret complex driving environments. Although many sensor modalities exist such as LiDAR, radar, and ultrasonic the automotive industry increasingly prioritizes camera based vision systems due to their lower cost and rich spatial detail. Typical AV systems include 6–10 cameras compared with 7 LiDAR units, resulting in substantial cost savings for large scale deployment. In this project, we examine how HDR (High Dynamic Range) and exposure settings in the camera image signal processing (ISP) pipeline affect the performance of a modern computer vision model YOLO-E when identifying objects in low light or highly variable lighting conditions. Our central research question is:		Autonomous vehicles rely heavily on robust perception systems to interpret complex driving environments. Although many sensor modalities exist such as LiDAR, radar, and ultrasonic the automotive industry increasingly prioritizes camera based vision systems due to their lower cost and rich spatial detail. Typical AV systems include 6–10 cameras compared with 7 LiDAR units, resulting in substantial cost savings for large scale deployment. In this project, we examine how HDR (High Dynamic Range) and exposure settings in the camera image signal processing (ISP) pipeline affect the performance of a modern computer vision model YOLO-E when identifying objects in low light or highly variable lighting conditions.

			Our central research question is:
	''How do different HDR and exposure settings influence YOLO’s confidence and reliability in low-light driving scenes?''		''How do different HDR and exposure settings influence YOLO’s confidence and reliability in low-light driving scenes?''

Graykim: /* Conclusions */

2025-12-09T20:06:09Z

Conclusions

← Older revision		Revision as of 20:06, 9 December 2025
Line 152:		Line 152:
	* Automotive HDR sensors consistently outperform consumer smartphone sensors for AV tasks.		* Automotive HDR sensors consistently outperform consumer smartphone sensors for AV tasks.
	* Split-pixel architectures (e.g., OV2312) provide the most robust overall performance.		* Split-pixel architectures (e.g., OV2312) provide the most robust overall performance.
			[[File:ResultsConfidence.png\|500px]]
			[[File:CombinedConfidenceIntervals.png\|650px]]
			[[File:SensorDiscussion.png\|500px]]

	== Appendix ==		== Appendix ==

	[[File:Snip 20210106183207.png\|200px]]		[[File:Snip 20210106183207.png\|200px]]

Graykim: /* YOLO */

2025-12-09T20:02:57Z

YOLO

← Older revision		Revision as of 20:02, 9 December 2025
Line 118:		Line 118:
	* Confidence interval		* Confidence interval
	Heatmaps are generated to visualize how confidence varies with exposure time across sensors.		Heatmaps are generated to visualize how confidence varies with exposure time across sensors.
			[[File:YoloSigmoid.png\|200px]]

	== Results ==		== Results ==