-
|
I am trying to record the paths of rays sampled in multiple directions, but I'm not quite sure how to do that. I thought I could start with the "Script a Renderer" tutorial, but there are a few things I don't understand:
Attaching pics of what I am trying to do, and the code block I currently have. Any help is very much appreciated! # Primary Surface Intersection
si = scene.ray_intersect(rays)
bsdf = si.bsdf(rays)
# Standard BSDF evaluation context for path tracing
bsdf_ctx = mi.BSDFContext()
# Getting Sampler from Camera
loaded_sensor = mi.load_dict(sensor)
sampler = loaded_sensor.sampler().clone()
# Loop iteration counter
i = mi.UInt32(0)
# Initialize the loop state (listing all variables that are modified inside the loop)
loop = mi.Loop(name="", state=lambda: (i, si,))
# while loop(si.is_valid() & (i < ambient_ray_count)):
with dr.suspend_grad():
# 1. Compute directions of initial surface interaction
bsdf_sample, bsdf_weight = bsdf.sample(bsdf_ctx, si, sampler.next_1d(),
sampler.next_2d())
# 2. Spawn a new ray starting at the surface interactions
rays2= si.spawn_ray(si.to_world(bsdf_sample.wo))
|
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 5 replies
-
|
Hi @h-OUS-e
Note: You'll most likely want to use a recorded loop (
|
Beta Was this translation helpful? Give feedback.
-
|
The for loop seems to be working and is generating the desired results, thank you! Attaching an image for reference. It is slow as you mentioned, and I wanted to see if mi.loop could be used. I don't quite understand though why my kernel crashes with the following code: # Loop iteration counter
i = mi.UInt32(0)
# Initialize the loop state (listing all variables that are modified inside the loop)
loop = mi.Loop(name="", state=lambda: (i, si, sampler))
while loop(si.is_valid() & (i < config['num_of_passes'])):
# 1. Get BSDF from Surface Interaction and context
try:
if len(ray_points) >= 0:
pass
except Exception:
ray_points = [ray_origins_list]
bsdf = si.bsdf(rays)
bsdf_ctx = mi.BSDFContext()
# 2. Sample new ray directions
bsdf_sample, bsdf_weight = bsdf.sample(bsdf_ctx, si, sampler.next_1d(),
sampler.next_2d())
# 3. Spawn a new ray starting at the surface interactions
rays2= si.spawn_ray(si.to_world(bsdf_sample.wo))
# 4. Record points
ray_points.append(np.array(rays2.o).copy().tolist())
# 5. Calculate new surface interactions
si = scene.ray_intersect(rays2)
# 6. Increase loop Iteration Count
i+=1The code runs just find without step 4, but introducing step 4 makes it crash for some reason. I am attaching the error below: |
Beta Was this translation helpful? Give feedback.
-
|
I think there's something about Windows & Jupyter that is hiding the DrJit error message. In short, recorded loops are just a recording of the computation they are not executed as the Python interpreter reaches them. Appending to the python list cannot be recorded, as it's an operation outside of the DrJit framework: we need to provide the list with an actual value. This requires DrJit to compute the value to be appended at every iteration of the loop. |
Beta Was this translation helpful? Give feedback.
-
|
I see, as a workaround, it seems like I got it to work by defining a class outside the loop, and calling dr.scatter to record ray origins, thank you! |
Beta Was this translation helpful? Give feedback.
-
|
Not sure if I should post this here, but on another note, when the light paths have no where to go, si.p returns [0,0,0] for that particular light path. How can I check if the light path has no where to go, and thus add an arbitrarily large number for that particular light path? Is there a way to do it without going through every single light path sequentially in a for loop? Trying to make the light paths not connect back to the origin point. |
Beta Was this translation helpful? Give feedback.
-
|
By "nowhere to go" I assume you're saying that they do not hit a single object in your scene? You can use |
Beta Was this translation helpful? Give feedback.
-
|
Thank you! |
Beta Was this translation helpful? Give feedback.
Hi @h-OUS-e
si.pto some normal Python list. In addition, of course, you'll want to save a mask to know which rays are still active.Note: You'll most likely want to use a recorded loop (
mi.Loop) for performance which is incompatible with what I mentioned above about visualizations. So you can either use a normal Python loop, or not save ray paths during optimization or even have a separate code path just for the visualizations.