3 ms·
Putting aside energy costs, Object detection is still crappy and has stalled. YOLO/SSDMN were impressive as all get-out, but they stink for general purpose use.
by SavantIdiot 5y ago
Putting aside energy costs, Object detection is still crappy and has stalled. YOLO/SSDMN were impressive as all get-out, but they stink for general purpose use. It's been 3 years (?) and general object detection, even with 100 classes, is still unusable off the shelf. Yes, I understand incremental training of pre-trained nets is a thing, but that's not where we all hoped it would go.
- thewarrior 5y agoIf object detection is crappy then how do they plan to make Tesla’s drive themselves?
- SavantIdiot 5y agoYou're on the right track and don't even know it! ;-P That's why the current Tesla has like Five LIDARs, two radars, 16 cameras, and lots of traditional algorithms that aren't NNets. (I don't know the counts, but my point stands.) I've spent 5 years on this with tier 1 automotives, I don't think ADAS5 will happen in my lifetime, and it definitely won't be due to neural nets: It is FAR more likely autonomy will be achieved by billions of IoT sensors in a mesh, guiding vehicles: e.g., lane sensors, weather sensors, c2c (car-to-car telemetry), c2i (car-to-infrastructure telemetry, like traffic lights), rather than some fancy in-car brain-emulation. Some of my favorite tesla fails: [1] https://techstory.in/tesla-autopilot-tricked-by-yellow-moon-into-thinking-yellow-traffic-light/ https://techstory.in/tesla-autopilot-tricked-by-yellow-moon-... [2] https://jalopnik.com/tesla-autopilot-glitch-makes-for-something-that-looks-l-1847024978 https://jalopnik.com/tesla-autopilot-glitch-makes-for-someth... [3] https://www.theregister.com/2019/04/02/tencent_tesla_hacking/ https://www.theregister.com/2019/04/02/tencent_tesla_hacking...
- thewarrior 5y agoVery interesting. But hasn’t Elon been promising even older Tesla owners a pure software upgrade to “Full Self Driving”