You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thanks for your great effort!
I have some doubts about the --scale parameter, I notice that when you benchmark ngp, you set scale=0.5 for default and 16 for 360v2 dataset, how does it help?
Also when I tried to set scale==1 for tanks&temple, the psnr dropped significantly (from 28 to 10), Are there some other parameters I should change?
Also, is scale bounded to the N_max of the hash grid? cause I found this line of code in networks.py?
"""
b = np.exp(np.log(2048scale/N_min)/(L-1))
"""
I believe N_max is 2048scale in this case and scale is from the --scale from the opt.py or command line right?
Looking forward to your response! Thanks Again!
The text was updated successfully, but these errors were encountered:
Thanks for your great effort!
I have some doubts about the --scale parameter, I notice that when you benchmark ngp, you set scale=0.5 for default and 16 for 360v2 dataset, how does it help?
Also when I tried to set scale==1 for tanks&temple, the psnr dropped significantly (from 28 to 10), Are there some other parameters I should change?
Also, is scale bounded to the N_max of the hash grid? cause I found this line of code in networks.py?
"""
b = np.exp(np.log(2048scale/N_min)/(L-1))
"""
I believe N_max is 2048scale in this case and scale is from the --scale from the opt.py or command line right?
Looking forward to your response! Thanks Again!
The text was updated successfully, but these errors were encountered: