search

agrimgupta92 / sgan

Code for "Social GAN: Socially Acceptable Trajectories with Generative Adversarial Networks", Gupta et al, CVPR 2018
MIT License
813 stars 261 forks source link

Question: Error with shape #32

Closed saruvora closed 5 years ago

saruvora commented 5 years ago

Hi,

I am evaluating eth pretrained model with my own dataset. I run into this:

curr_seq[_idx, :, pad_front:pad_end] = curr_ped_seq ValueError: could not broadcast input array from shape (2,44) into shape (2,16)

error. I am not able to understand if I have to change any parameters or not while testing with my own data. Can you please help me out ?

atanas1054 commented 5 years ago

I am having the same problem. Any ideas?

saruvora commented 5 years ago

yes, I solved it. The dataset that you are using may be has more number of sequences than desired by trajectories.py if pad_end - pad_front != self.seq_len: continue

this part causes the error. So just check how many points does each pedestrian has. You'll need it less than the sequence length. And also check for the similar repeated points of each pedestrian(x,y). The dataset used has different for each frame. After you solve this you may have errors if you solve them let me know.

atanas1054 commented 5 years ago

Got it. Thank you.