idstein / adam Goto Github PK

Hi Paul,
Sorry to raise an issue here, I had some doubts regarding enabling MACE OpenCL backend on Tflite.
I found a comment of yours on the tensorflow repo that said using MACE as GPU backend improved around 30% speedup of execution.
Can you show me how to enable the same for GPU pipeline or show me some repo where this has been done?
Thanks a lot for your help :)

Refer to http://www.cvlibs.net/datasets/kitti/eval_road.php for benchmark results.

https://github.com/jmedveckyh/Hough

Evaluate how to show curvy directions
Maybe use accelerometer to detect centrifugal force

For test sets refer to http://benchmark.ini.rub.de

https://github.com/Akson/LaneDepartureWarning

We should use GPS based as it is almost accurate with speeds over 33 Km/h.

Otherwise we must implement an auto-calibration for the camera. In addition to that, it may fail as the world position of the camera as well as it angles may vary due to invariants in the street structure.

• Define acceptance tests
• Define success / failure criteria
• Collect test sets (e.g. https://github.com/Alexoner/vehicle-assistance, http://www.cvlibs.net/datasets/kitti/eval_road.php, http://benchmark.ini.rub.de)

• We may need the rough distance between our car and the car in front of us.
• We need to continuously calibrate our camera to have a real-world position/intrinsic parameters (as focal point may change)