Open sahilm1992 opened 4 years ago
In the training stage, we do correct with the ground truth road segments.
Never explicitly checking whether runs into dead loop, but we use the max_length
parameter to guarantee the termination of the generation process.
Thanks.
Do you have any stats on how many times did you reach the destination?
Actually, this paper which you have compared with "https://www.ijcai.org/Proceedings/2017/0430.pdf" is correcting itself during test time as far as I understand. Have you used their code in your implementation?
Could you please update me on these things. Thanks
If for a test trajectory, the route is A B C D E F G H I J let us say current route predicted till time ti is A B C D
What if next node predicted is a node Q instead of E? Do you correct it to E or keep it Q and keep running the algorithm?
Also,do you ever encounter dead ends?
Could you please clarify this.