Struct libredr::PyLibreDR

source ·

pub struct PyLibreDR(pub(crate) LibreDR, pub(crate) String, pub(crate) bool, pub(crate) bool);

Expand description

Python interface for LibreDR client

Tuple Fields§

§0: LibreDR§1: String§2: bool§3: bool

Implementations§

source §

pub fn py_new( py: Python<'_>, connect: String, unix: bool, tls: bool, ) -> Result<Self>

Construct LibreDR by connecting to LibreDR server.

See LibreDR::new for details.

source

pub fn getnewargs(&self) -> (String, bool, bool)

To allow pickle PyLibreDR object by reconnecting to the server.

Unpickled connection has different UUID.

source

pub fn py_ray_tracing_forward<'py>( &mut self, py: Python<'py>, geometry: &PyGeometry, ray: PyReadonlyArrayDyn<'_, f32>, texture: PyReadonlyArray3<'_, f32>, envmap: PyReadonlyArray4<'_, f32>, sample_per_pixel: Py<PyAny>, max_bounce: Py<PyAny>, switches: (u8, u8, u8, u8), clip_near: Py<PyAny>, camera_space: bool, requires_grad: bool, srand: i32, low_discrepancy: Option<u32>, ) -> Result<Bound<'py, PyArrayDyn<f32>>>

Create a RequestRayTracingForward task and wait for response

§Arguments

ray - ray parameters
- if camera_space is false 18 * image_shape
 - including ray position 9 * image_shape
 - including ray direction 9 * image_shape
- if camera_space is true, add another (1 + 14) channels
 - including ray depth 1 * image_shape (if depth <= 0, treat as hit miss)
 - including ray material 14 * image_shape
texture - (3 + 3 + 3 + 1 + 3 + 1) * texture_resolution * texture_resolution (must be square image)
- including normal + diffuse + specular + roughness + intensity + window
envmap - 3 * 6 * envmap_resolution * envmap_resolution
- (must be box unwrapped 6 square images)
sample_per_pixel - sample_per_pixel_forward, (sample_per_pixel_backward)
- sample_per_pixel can be a single integer,
  - (same value for forward and backward)
- or tuple of 2 integers.
  - (only sample_per_pixel_backward number of rays are stored for backward)
  - (must ensure sample_per_pixel_forward >= sample_per_pixel_backward)
max_bounce - max_bounce_forward, (max_bounce_backward), (max_bounce_low_discrepancy), (skip_bounce)
- max_bounce can be a single integer, or tuple of 2-4 integers.
- The default value for max_bounce_backward is the same as max_bounce_forward.
- The default value for max_bounce_low_discrepancy is 0.
- The default value for skip_bounce is 0.
switches - tuple of 4 switches to determine hit miss and reflection behavior
- render::MISS_* - determine how to deal with ray hit miss
  - common::render::MISS_NONE
  - common::render::MISS_ENVMAP
- render::REFLECTION_NORMAL_* - determine how to get surface normal
- render::REFLECTION_DIFFUSE_* - determine diffuse reflection model
  - common::render::REFLECTION_DIFFUSE_NONE
  - common::render::REFLECTION_DIFFUSE_LAMBERTIAN
- render::REFLECTION_SPECULAR_* - determine specular reflection model
clip_near - clip near distance for camera
- clip_near can be a single float number (same for all bounces),
- or tuple of 3 float numbers (first bounce, second bounce, and other bounces)
camera_space - if true, the first bounce uses the depth and material given by the ray
requires_grad - if true, worker will save intermediate data, the next task must be ray_tracing_backward
srand - random seed
- if srand >= 0, the same random seed is used for every pixel
- if srand < 0, use different seed for each pixel
low_discrepancy - (optional) start id of Halton low discrepancy sequence.
- The default value is the same as sample_per_pixel_forward.
- if combine multiple rendered images to reduce noise, this value can be set to:
  1 * sample_per_pixel_forward, 2 * sample_per_pixel_forward, 3 * sample_per_pixel_forward, …

§Return

Return shape will be,

if camera_space is true
- render image 3 * image_shape
if camera_space is false, add another
- ray texture coordinate 2 * image_shape
- ray depth (Euclidean distance) 1 * image_shape
- ray normal 3 * image_shape

source

pub fn py_ray_tracing_backward<'py>( &mut self, py: Python<'py>, d_ray: PyReadonlyArrayDyn<'_, f32>, ) -> Result<(Bound<'py, PyArray3<f32>>, Bound<'py, PyArray4<f32>>, Option<Bound<'py, PyArrayDyn<f32>>>)>

Create a RequestRayTracingBackward task and wait for response.

Must be called consecutive to a RequestRayTracingForward task with requires_grad set to true.
To create multiple RequestRayTracingForward tasks and backward together, multiple client connections are required.

§Arguments

d_ray - gradient of image 3 * image_shape (must ensure same image_shape as RequestRayTracingForward)

§Return

Return shape will be,

if camera_space is false for RequestRayTracingForward task
- 1st return value (3 + 3 + 3 + 1 + 3 + 1) * texture_resolution * texture_resolution
  - (same texture_resolution as RequestRayTracingForward)
  - including d_normal + d_diffuse + d_specular + d_roughness + d_intensity + d_window
- 2nd return value 3 * 6 * envmap_resolution * envmap_resolution
  - (same envmap_resolution as RequestRayTracingForward)
  - including d_envmap
if camera_space is true for RequestRayTracingForward task, add another
- 3rd return value 14 * image_shape (same shape as RequestRayTracingForward)
  - including d_ray_texture