lmb-freiburg / cv-exercises Goto Github PK

In exercise sheet we are told to calculate the average endpoint error and pointwise endpoint error. I had looked it up online and the method to calculate endpoint error was given as norm(Vestemated - Vground_truth)[1]. According to the function available to calculate the pointwise endpoint error in the template it can be calculated using [1]. But the function for average endpoint error has epsilon as an argument in the metrics.py. How are we supposed to calculate the average? are we supposed to average it with the size of estimated flow. I am a little stuck in this.

While calculating the photometric loss we need the flow vectors (u,v), the template in the photometricloss class in loss.py doesn't have any information about the vectors, it only has sample and model_output as arguments, am I missing anything in this.

Could you kindly help me regarding this query.