I've been surprised at how many things 10.69 visualizes with the gray blob for occupied driveable space. Perhaps part of it is just driving manually, one just naturally avoids objects without much thought. Also kinda interesting is how occupancy is visualized near cones and trash containers as it's pretty clear the displayed graphics is from the old old "FSD Visualization Preview" with only approximately placed visualization, and FSD Beta happily drives through these misplaced objects; whereas the occupancy network might have the highest accuracy placement of these resulting in blobs in the "true" location.
Just in the last day, I've seen it correctly visualize and drive around construction signs (e.g., folding A-frame barricades, orange diamond work zone signs, folded-down/now-triangle signs), tables placed behind a work van, random stuff from a yard sale, electric scooter fallen down from the sidewalk, yard clippings, fallen branches, and more.
With so much variety and relatively high frequency -- much more often than taking an unprotected left turn, I wonder if Tesla prioritized the occupancy network for high disengagement rates to avoid running into these things?
Just in the last day, I've seen it correctly visualize and drive around construction signs (e.g., folding A-frame barricades, orange diamond work zone signs, folded-down/now-triangle signs), tables placed behind a work van, random stuff from a yard sale, electric scooter fallen down from the sidewalk, yard clippings, fallen branches, and more.
With so much variety and relatively high frequency -- much more often than taking an unprotected left turn, I wonder if Tesla prioritized the occupancy network for high disengagement rates to avoid running into these things?