In this talk, Dave will introduce optimization, computational imaging, and computer vision approaches being successfully used to handle such edge cases. For teams in the architecture phase and able to modify their perception stack for the best possible performance, he will showcase a way to revisit the camera design itself and co-optimize new processing stacks, from sensing to detection. For teams unable to change the vision system architecture of their current or in-production designs, he will show how automated optimization of the camera imaging pipeline can quickly improve computer vision results by significant margins. I will share results from a case study from an automotive Tier 1 program that delivered over 25% mAP improvement in object detection performance vs. their current methodology.