Add the rest of C++ ray tracing implementation (#3630)

mthrok · web-flow · commit ff266b153959 · 2023-10-02T16:33:17.000-04:00
Taken from #3234
diff --git a/torchaudio/csrc/CMakeLists.txt b/torchaudio/csrc/CMakeLists.txt
@@ -42,7 +42,7 @@ if(BUILD_RNNT)
 endif()
 
 if(BUILD_RIR)
-  list(APPEND sources rir/rir.cpp)
+  list(APPEND sources rir/rir.cpp rir/ray_tracing.cpp)
   list(APPEND compile_definitions INCLUDE_RIR)
 endif()
 
diff --git a/torchaudio/csrc/rir/ray_tracing.cpp b/torchaudio/csrc/rir/ray_tracing.cpp
@@ -0,0 +1,374 @@
+/*
+Copyright (c) 2014-2017 EPFL-LCAV
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+*/
+
+//
+// Ray tracing implementation. This is heavily based on PyRoomAcoustics:
+// https://github.com/LCAV/pyroomacoustics
+//
+#include <torch/script.h>
+#include <torch/torch.h>
+#include <torchaudio/csrc/rir/wall.h>
+#include <cmath>
+
+namespace torchaudio {
+namespace rir {
+namespace {
+
+// TODO: remove this once hybrid method is supported
+const bool IS_HYBRID_SIM = false;
+
+// TODO: remove this once ISM method is supported
+const int ISM_ORDER = 10;
+
+#define EPS ((scalar_t)(1e-5))
+#define VAL(x) ((x).template item<scalar_t>())
+#define NORM(x) (VAL((x).norm()))
+#define MAX(x) (VAL((x).max()))
+#define IN_RANGE(x, y) ((-EPS < (x)) && ((x) < (y) + EPS))
+
+template <typename scalar_t, unsigned int D>
+const std::array<Wall<scalar_t>, D * 2> make_walls(
+    const torch::Tensor& room,
+    const torch::Tensor& absorption,
+    const torch::Tensor& scattering) {
+  if constexpr (D == 2) {
+    auto w = room.index({0}).item<scalar_t>();
+    auto l = room.index({1}).item<scalar_t>();
+    return make_room<scalar_t>(w, l, absorption, scattering);
+  }
+  if constexpr (D == 3) {
+    auto w = room.index({0}).item<scalar_t>();
+    auto l = room.index({1}).item<scalar_t>();
+    auto h = room.index({2}).item<scalar_t>();
+    return make_room<scalar_t>(w, l, h, absorption, scattering);
+  }
+}
+
+inline double get_energy_coeff(
+    const double travel_dist,
+    const double mic_radius_sq) {
+  double sq = travel_dist * travel_dist;
+  auto p_hit = 1. - std::sqrt(1. - mic_radius_sq / std::max(mic_radius_sq, sq));
+  return sq * p_hit;
+}
+
+/// RayTracer class helper for ray tracing.
+/// For attribute description, Python wrapper.
+template <typename scalar_t, unsigned int D>
+class RayTracer {
+  // Provided parameters
+  const torch::Tensor& room;
+  const torch::Tensor& mic_array;
+  const double mic_radius;
+
+  // Values derived from the parameters
+  const int num_bands;
+  const double mic_radius_sq;
+  const bool do_scattering; // Whether scattering is needed (scattering != 0)
+  const std::array<Wall<scalar_t>, D * 2> walls; // The walls of the room
+
+  // Runtime value caches
+  // Updated at the beginning of the simulation
+  double sound_speed = 343.0;
+  double distance_thres = 10.0 * sound_speed; // upper bound
+  double energy_thres = 0.0; // lower bound
+  double hist_bin_width = 0.004; // [second]
+
+ public:
+  RayTracer(
+      const torch::Tensor& room,
+      const torch::Tensor& absorption,
+      const torch::Tensor& scattering,
+      const torch::Tensor& mic_array,
+      const double mic_radius)
+      : room(room),
+        mic_array(mic_array),
+        mic_radius(mic_radius),
+        num_bands(absorption.size(0)),
+        mic_radius_sq(mic_radius * mic_radius),
+        do_scattering(MAX(scattering) > 0.),
+        walls(make_walls<scalar_t, D>(room, absorption, scattering)) {}
+
+  // The main (and only) public entry point of this class. The histograms Tensor
+  // reference is passed along and modified in the subsequent private method
+  // calls. This method spawns num_rays rays in all directions from the source
+  // and calls simul_ray() on each of them.
+  torch::Tensor compute_histograms(
+      const torch::Tensor& origin,
+      int num_rays,
+      double time_thres,
+      double energy_thres_ratio,
+      double sound_speed_,
+      int num_bins) {
+    scalar_t energy_0 = 2. / num_rays;
+    auto energies = torch::full({num_bands}, energy_0, room.options());
+
+    auto histograms =
+        torch::zeros({mic_array.size(0), num_bins, num_bands}, room.options());
+
+    // Cache runtime parameters
+    sound_speed = sound_speed_;
+    energy_thres = energy_0 * energy_thres_ratio;
+    distance_thres = time_thres * sound_speed;
+    hist_bin_width = time_thres / num_bins;
+
+    // TODO: the for loop can be parallelized over num_rays by creating
+    // `num_threads` histograms and then sum-reducing them into a single
+    // histogram.
+    static_assert(D == 2 || D == 3, "Only 2D and 3D are supported.");
+    if constexpr (D == 2) {
+      scalar_t delta = 2. * M_PI / num_rays;
+      for (int i = 0; i < num_rays; ++i) {
+        scalar_t phi = i * delta;
+        auto dir = torch::tensor({cos(phi), sin(phi)}, room.scalar_type());
+        simul_ray(energies, origin, dir, histograms);
+      }
+    } else {
+      scalar_t delta = 2. / num_rays;
+      scalar_t increment = M_PI * (3. - std::sqrt(5.)); // phi increment
+
+      for (auto i = 0; i < num_rays; ++i) {
+        auto z = (i * delta - 1) + delta / 2.;
+        auto rho = std::sqrt(1. - z * z);
+
+        scalar_t phi = i * increment;
+
+        auto x = cos(phi) * rho;
+        auto y = sin(phi) * rho;
+
+        auto azimuth = atan2(y, x);
+        auto colatitude = atan2(std::sqrt(x * x + y * y), z);
+
+        auto dir = torch::tensor(
+            {sin(colatitude) * cos(azimuth),
+             sin(colatitude) * sin(azimuth),
+             cos(colatitude)},
+            room.scalar_type());
+
+        simul_ray(energies, origin, dir, histograms);
+      }
+    }
+    return histograms.transpose(1, 2); // (num_mics, num_bands, num_bins)
+  }
+
+ private:
+  /// Get the bin index from the distance traveled to a mic.
+  inline int get_bin_idx(scalar_t travel_dist_at_mic) {
+    auto time_at_mic = travel_dist_at_mic / sound_speed;
+    return (int)floor(time_at_mic / hist_bin_width);
+  }
+
+  ///
+  /// Traces a single ray. phi (horizontal) and theta (vectorical) are the
+  /// angles of the ray from the source. Theta is 0 for 2D rooms.  When a ray
+  /// intersects a wall, it is reflected and part of its energy is absorbed. It
+  /// is also scattered (sent directly to the microphone(s)) according to the
+  /// scattering coefficient. When a ray is close to the microphone, its current
+  /// energy is recoreded in the output histogram for that given time slot.
+  ///
+  /// See also:
+  /// https://github.com/LCAV/pyroomacoustics/blob/df8af24c88a87b5d51c6123087cd3cd2d361286a/pyroomacoustics/libroom_src/room.cpp#L855-L986
+  void simul_ray(
+      torch::Tensor& energies,
+      torch::Tensor origin,
+      torch::Tensor dir,
+      torch::Tensor& histograms) {
+    auto travel_dist = 0.;
+    // To count the number of times the ray bounces on the walls
+    // For hybrid generation we add a ray to output only if specular_counter
+    // is higher than the ism order.
+    int specular_counter = 0;
+    while (true) {
+      // Find the next hit point
+      auto [hit_point, next_wall_index, hit_distance] =
+          find_collision_wall<scalar_t, D>(room, origin, dir);
+
+      auto& wall = walls[next_wall_index];
+
+      // Check if the specular ray hits any of the microphone
+      if (!(IS_HYBRID_SIM && specular_counter < ISM_ORDER)) {
+        // Compute the distance between the line defined by (origin, hit_point)
+        // and the center of the microphone (mic_pos)
+
+        for (auto mic_idx = 0; mic_idx < mic_array.size(0); mic_idx++) {
+          //
+          //                 _  o microphone
+          //         to_mic  / |   ^
+          //               /       |              wall
+          //             /         | mic radious  | |
+          //   origin  /           |              | |
+          //         /             v              | |
+          //       x ---------------------------> |x| collision
+          //
+          //       | <--------> |
+          //       impact_distance
+          //       | <--------------------------> |
+          //               hit_distance
+          //
+          torch::Tensor to_mic = mic_array[mic_idx] - origin;
+          scalar_t impact_distance = VAL(to_mic.dot(dir));
+
+          // mic is further than the collision point.
+          // So microphone did not pick up the sound.
+          if (!IN_RANGE(impact_distance, hit_distance)) {
+            continue;
+          }
+
+          // If the ray hit the coverage of the mic, compute the energy
+          if (NORM(to_mic - dir * impact_distance) < mic_radius + EPS) {
+            // The length of this last hop
+            auto travel_dist_at_mic = travel_dist + std::abs(impact_distance);
+            auto coeff = get_energy_coeff(travel_dist_at_mic, mic_radius_sq);
+            auto energy = energies / coeff;
+            histograms[mic_idx][get_bin_idx(travel_dist_at_mic)] += energy;
+          }
+        }
+      }
+
+      travel_dist += hit_distance;
+      energies *= wall.reflection;
+
+      //   Let's shoot the scattered ray induced by the rebound on the wall
+      if (do_scattering) {
+        scat_ray(histograms, wall, energies, origin, hit_point, travel_dist);
+        energies *= (1. - wall.scattering);
+      }
+
+      // Check if we reach the thresholds for this ray
+      if (travel_dist > distance_thres || VAL(energies.max()) < energy_thres) {
+        break;
+      }
+
+      // set up for next iteration
+      specular_counter += 1;
+      dir = reflect(wall, dir);
+      origin = hit_point;
+    }
+  }
+
+  ///
+  /// Scatters a ray towards the microphone(s), i.e. records its scattered
+  /// energy in the histogram. Called when a ray hits a wall.
+  ///
+  /// See also:
+  /// https://github.com/LCAV/pyroomacoustics/blob/df8af24c88a87b5d51c6123087cd3cd2d361286a/pyroomacoustics/libroom_src/room.cpp#L761-L853
+  void scat_ray(
+      torch::Tensor& histograms,
+      const Wall<scalar_t>& wall,
+      const torch::Tensor& energies,
+      const torch::Tensor& prev_hit_point,
+      const torch::Tensor& hit_point,
+      scalar_t travel_dist) {
+    for (auto mic_idx = 0; mic_idx < mic_array.size(0); mic_idx++) {
+      auto mic_pos = mic_array[mic_idx];
+      if (side(wall, mic_pos) != side(wall, prev_hit_point)) {
+        continue;
+      }
+
+      // As the ray is shot towards the microphone center,
+      // the hop dist can be easily computed
+      torch::Tensor hit_point_to_mic = mic_pos - hit_point;
+      auto hop_dist = NORM(hit_point_to_mic);
+      auto travel_dist_at_mic = travel_dist + hop_dist;
+
+      // compute the scattered energy reaching the microphone
+      auto h_sq = hop_dist * hop_dist;
+      auto p_hit_equal = 1. - std::sqrt(1. - mic_radius_sq / h_sq);
+      // cosine angle should be positive, but could be negative if normal is
+      // facing out of room so we take abs
+      auto p_lambert = (scalar_t)2. * std::abs(cosine(wall, hit_point_to_mic));
+      auto scat_trans = wall.scattering * energies * p_hit_equal * p_lambert;
+
+      if (travel_dist_at_mic < distance_thres &&
+          MAX(scat_trans) > energy_thres) {
+        auto coeff = get_energy_coeff(travel_dist_at_mic, mic_radius_sq);
+        auto energy = scat_trans / coeff;
+        histograms[mic_idx][get_bin_idx(travel_dist_at_mic)] += energy;
+      }
+    }
+  }
+};
+
+///
+/// @brief Compute energy histogram via ray tracing. See Python wrapper for
+/// detail about parameters and output.
+///
+torch::Tensor ray_tracing(
+    const torch::Tensor& room,
+    const torch::Tensor& source,
+    const torch::Tensor& mic_array,
+    int64_t num_rays,
+    const torch::Tensor& absorption,
+    const torch::Tensor& scattering,
+    double mic_radius,
+    double sound_speed,
+    double energy_thres,
+    double time_thres, // TODO: rename to duration
+    double hist_bin_size) {
+  // TODO: Raise this to Python layer
+  auto num_bins = (int)ceil(time_thres / hist_bin_size);
+  switch (room.size(0)) {
+    case 2: {
+      return AT_DISPATCH_FLOATING_TYPES(
+          room.scalar_type(), "ray_tracing_2d", [&] {
+            RayTracer<scalar_t, 2> rt(
+                room, mic_array, absorption, scattering, mic_radius);
+            return rt.compute_histograms(
+                source,
+                num_rays,
+                time_thres,
+                energy_thres,
+                sound_speed,
+                num_bins);
+          });
+    }
+    case 3: {
+      return AT_DISPATCH_FLOATING_TYPES(
+          room.scalar_type(), "ray_tracing_3d", [&] {
+            RayTracer<scalar_t, 3> rt(
+                room, mic_array, absorption, scattering, mic_radius);
+            return rt.compute_histograms(
+                source,
+                num_rays,
+                time_thres,
+                energy_thres,
+                sound_speed,
+                num_bins);
+          });
+    }
+    default:
+      TORCH_CHECK(false, "Only 2D and 3D are supported.");
+  }
+}
+
+TORCH_LIBRARY_IMPL(torchaudio, CPU, m) {
+  m.impl("torchaudio::ray_tracing", torchaudio::rir::ray_tracing);
+}
+
+TORCH_LIBRARY_FRAGMENT(torchaudio, m) {
+  m.def(
+      "torchaudio::ray_tracing(Tensor room, Tensor source, Tensor mic_array, int num_rays, Tensor absorption, Tensor scattering, float mic_radius, float sound_speed, float energy_thres, float time_thres, float hist_bin_size) -> Tensor");
+}
+
+} // namespace
+} // namespace rir
+} // namespace torchaudio
diff --git a/torchaudio/csrc/rir/wall.h b/torchaudio/csrc/rir/wall.h
