Presented: Dijing Zhang
visualize grid
visualize rays
sampled points along the rays
feature visualization
depth visualization
Report the center of the box, and the side lengths of the box after training, rounded to the nearest 1/100 decimal place.
Box center: (0.25021374225616455, 0.25057095289230347, -0.000454530818387866)
Box side lengths: (2.005078077316284, 1.5035607814788818, 1.5033352375030518)
Discuss the trade-offs between increased view dependence and generalization quality.
For original NeRF model, it takes view direction and positions as input and outputs the density and feature. So given the view direciton, it can provide a more detailed result. But meanwhile, it may well cause overfitting since it give a specifict direciton. It is quite important to blanace the weight of view directions, in order to achieve a genralized and good result.