Hey @MankaranSingh , thanks for your continuous effort with the chapter!
I do not want to merge this PR, since I do not like the changes to calibrated_lane_detector.py. Some of the improvements I had made with my most recent commits are being reverted with this PR. These improvements are
very little code (also got rid of plotting code). makes it faster and easier to understand
got rid of calibrate method, and doing everything in __call__. In the previous design the user would need to call calibrateand then __call__, if he/she wants calibration and lane detection. But with this way, the detect method gets called twice and hence the neural net gets executed twice, which is very bad for runtime. The new implementation solves this.
Hey @MankaranSingh , thanks for your continuous effort with the chapter!
I do not want to merge this PR, since I do not like the changes to
calibrated_lane_detector.py
. Some of the improvements I had made with my most recent commits are being reverted with this PR. These improvements arecalibrate
method, and doing everything in__call__
. In the previous design the user would need to callcalibrate
and then__call__
, if he/she wants calibration and lane detection. But with this way, thedetect
method gets called twice and hence the neural net gets executed twice, which is very bad for runtime. The new implementation solves this.