..

Author: Bharath2

README.md

@@ -1,9 +1,11 @@
 # NeRF.cpp
+NeRF implementation with Siren layers in C++.
 
+| <img src="./output/rgb.gif" alt="NeRF RGB Animation" style="display:inline-block"/> | <img src="./output/depth.gif" alt="NeRF Depth Map" style="display:inline-block"/> |
+|:-------------:|:---------:|
+| **RGB** | **Depth Map** |
 
-<img src="./output/animation.gif" alt="NeRF Animation"/>
 
-NeRF implementation with Siren layers in C++.
 ## Requirements
 - CUDA
 - LibTorch

output/animation.gif

@@ -14,8 +14,8 @@ constexpr int n_final_frames = 30;
 //Render options
 constexpr int batch_size = 1280000;
 constexpr float start_distance = 2.0f;
-constexpr float end_distance = 8.0f;
-constexpr int n_samples = 96; //no. of samples along each Ray
+constexpr float end_distance = 6.0f;
+constexpr int n_samples = 128; //no. of samples along each Ray
 
 int main(int argc, char *argv[]) {
   // Parse command-line arguments
@@ -49,7 +49,7 @@ int main(int argc, char *argv[]) {
   std::uniform_int_distribution<int> dist(0, dataset.len - 1);
 
   // Train the NeRF model
-  for (int i = 0; i < n_iters; i++) {
+  for (int i = 0; i <= n_iters; i++) {
     // Sample a random image and its corresponding pose
     int img_i = dist(rng);
     auto target = dataset.images[img_i].to(device);