Antonio Loquercio

You are allowed to use the goal location and the entire state of the drone for training and evaluation. The final evaluation will be on depth images + states.

@gcioffi ptal

Could you maybe double check that the data and pipeline is alright? A possibility would be to overfit to that dataset. Then, maybe analyze what are the places with higher...

Hi @AHzZ123, Sorry for the poor naming convention. The function should be correct though

I guess you have a problem with the data processing pipeline. Could you follow the instruction in the repo? It should then work out of the box

I am sorry but I did not fully understand your question. Just to remind you, it is important to pre-process the data before the training with the procedure we outline...

I am sorry but we don't plan to release the full dataset (videos), as they have to undergo a large pre-processing before rewriting. I would recommend to collect your own...

I am extremely sorry but we won't be able to release the full video. I understand that this could be very important for you, but unfortunately we are not allowed...

Hi Dong thanks for your message. The open-cv used for drone control is the cv2bridge available for ROS and not python-opencv. As you correctly noticed, this is simpy a version...

Hi @yhkim0566, I think that you're missing the data pre-processing step. Therefore, the model output are not good.