search

zhou13 / neurvps

Neural Vanishing Point Scanning via Conic Convolution
MIT License
178 stars 22 forks source link

the train data about the model #22

Closed zyg11 closed 2 years ago

zyg11 commented 2 years ago

Hi zhou, Thanks for sharing your code. I tested the model in the laneline dataset , the performance was better than a lot algorithms. I wander whether the number of train data is just 2000. Because the performance is amazing.

zhou13 commented 2 years ago

Yes, and you should be able to reproduce it from scratch using the released code. (I think you are referring to the tmm17 model).