Remaining pieces for pca shape scalar

akenmorris · akenmorris · commit 7e2cf172ed71 · 2023-11-20T23:23:23.000-07:00
diff --git a/Python/shapeworks/shapeworks/shape_scalars.py b/Python/shapeworks/shapeworks/shape_scalars.py
@@ -90,7 +90,7 @@ def run_find_num_components(x, y, max_components, cv=5):
     return figdata_png
 
 
-def pred_from_mbpls(new_x, n_components=3):
+def pred_from_mbpls(new_x):
     """ Predict new_y from new_x using existing mbpls fit """
 
     if not does_mbpls_model_exist():
@@ -112,4 +112,16 @@ def does_mbpls_model_exist():
     except NameError:
         return False
 
-    return True
+    return True
+
+def clear_mbpls_model():
+    """ Clear mbpls model """
+
+    global mbpls_model
+    try:
+        mbpls_model
+    except NameError:
+        return
+
+    del mbpls_model
+    return
diff --git a/Studio/Analysis/AnalysisTool.cpp b/Studio/Analysis/AnalysisTool.cpp
@@ -463,6 +463,23 @@ void AnalysisTool::network_analysis_clicked() {
   app_->get_py_worker()->run_job(network_analysis_job_);
 }
 
+//-----------------------------------------------------------------------------
+Eigen::VectorXd AnalysisTool::extract_positions(Eigen::VectorXd& data) {
+  /*
+    auto positions = data;
+
+    if (pca_shape_plus_scalar_mode()) {
+      positions = extract_shape_only(data);
+    } else if (pca_scalar_only_mode()) {
+      computed_scalars_ = temp_shape_;
+      if (ui_->pca_predict_shape->isChecked()) {
+        positions = ShapeScalarJob::predict_shape(session_, QString::fromStdString(feature_map_), computed_scalars_);
+      } else {
+        positions = construct_mean_shape();
+      }
+    }*/
+}
+
 //-----------------------------------------------------------------------------
 bool AnalysisTool::compute_stats() {
   if (stats_ready_) {
@@ -637,6 +654,10 @@ Particles AnalysisTool::get_mean_shape_points() {
     return Particles();
   }
 
+  if (ui_->pca_scalar_only->isChecked()) {
+    return convert_from_combined(construct_mean_shape());
+  }
+
   if (ui_->group1_button->isChecked() || ui_->difference_button->isChecked()) {
     return convert_from_combined(stats_.get_group1_mean());
   } else if (ui_->group2_button->isChecked()) {
@@ -700,7 +721,11 @@ Particles AnalysisTool::get_shape_points(int mode, double value) {
     computed_scalars_ = extract_scalar_only(temp_shape_);
   } else if (pca_scalar_only_mode()) {
     computed_scalars_ = temp_shape_;
-    positions = construct_mean_shape();
+    if (ui_->pca_predict_shape->isChecked()) {
+      positions = ShapeScalarJob::predict_shape(session_, QString::fromStdString(feature_map_), computed_scalars_);
+    } else {
+      positions = construct_mean_shape();
+    }
   }
 
   return convert_from_combined(positions);
@@ -1245,7 +1270,7 @@ ShapeHandle AnalysisTool::create_shape_from_points(Particles points) {
   shape->set_reconstruction_transforms(reconstruction_transforms_);
 
   if (feature_map_ != "") {
-    if (ui_->pca_predict_scalar->isChecked()) {
+    if (ui_->pca_scalar_shape_only->isChecked() && ui_->pca_predict_scalar->isChecked()) {
       auto scalars = ShapeScalarJob::predict_scalars(session_, QString::fromStdString(feature_map_),
                                                      points.get_combined_global_particles());
       shape->set_point_features(feature_map_, scalars);
@@ -1730,6 +1755,11 @@ void AnalysisTool::change_pca_analysis_type() {
   stats_ready_ = false;
   evals_ready_ = false;
   stats_ = ParticleShapeStatistics();
+  ShapeScalarJob::clear_model();
+
+  ui_->pca_predict_scalar->setEnabled(ui_->pca_scalar_shape_only->isChecked());
+  ui_->pca_predict_shape->setEnabled(ui_->pca_scalar_only->isChecked());
+
   compute_stats();
   Q_EMIT pca_update();
 }
diff --git a/Studio/Analysis/AnalysisTool.h b/Studio/Analysis/AnalysisTool.h
@@ -194,9 +194,14 @@ class AnalysisTool : public QWidget {
 
   void change_pca_analysis_type();
 
+  //Eigen::VectorXd get_mean_shape();
+
   //! Compute the mean shape outside of the PCA in case we are using scalars only
   Eigen::VectorXd construct_mean_shape();
 
+
+  Eigen::VectorXd extract_positions(Eigen::VectorXd& data);
+
  Q_SIGNALS:
 
   void update_view();
diff --git a/Studio/Analysis/AnalysisTool.ui b/Studio/Analysis/AnalysisTool.ui
@@ -1451,14 +1451,14 @@ QWidget#particles_panel {
                       <item row="1" column="1">
                        <widget class="QCheckBox" name="pca_predict_scalar">
                         <property name="text">
-                         <string>Predict scalar</string>
+                         <string>Predict Scalar</string>
                         </property>
                        </widget>
                       </item>
                       <item row="2" column="1">
                        <widget class="QCheckBox" name="pca_predict_shape">
                         <property name="text">
-                         <string>Predict shape</string>
+                         <string>Predict Shape</string>
                         </property>
                        </widget>
                       </item>
diff --git a/Studio/Job/ShapeScalarJob.cpp b/Studio/Job/ShapeScalarJob.cpp
@@ -16,10 +16,12 @@ using namespace pybind11::literals;  // to bring in the `_a` literal
 
 namespace shapeworks {
 
+std::atomic<bool> ShapeScalarJob::needs_clear_ = false;
+
 //---------------------------------------------------------------------------
 ShapeScalarJob::ShapeScalarJob(QSharedPointer<Session> session, QString target_feature,
                                Eigen::MatrixXd target_particles, JobType job_type)
-    : session_(session), target_feature_(target_feature), target_particles_(target_particles), job_type_(job_type) {}
+    : session_(session), target_feature_(target_feature), target_values_(target_particles), job_type_(job_type) {}
 
 //---------------------------------------------------------------------------
 void ShapeScalarJob::run() {
@@ -68,10 +70,23 @@ QPixmap ShapeScalarJob::get_plot() { return plot_; }
 //---------------------------------------------------------------------------
 Eigen::VectorXd ShapeScalarJob::predict_scalars(QSharedPointer<Session> session, QString target_feature,
                                                 Eigen::MatrixXd target_particles) {
+  return predict(session, target_feature, target_particles, Direction::To_Scalar);
+}
+
+//---------------------------------------------------------------------------
+Eigen::VectorXd ShapeScalarJob::predict_shape(QSharedPointer<Session> session, QString target_feature,
+                                              Eigen::MatrixXd target_scalars) {
+  return predict(session, target_feature, target_scalars, Direction::To_Shape);
+}
+
+//---------------------------------------------------------------------------
+Eigen::VectorXd ShapeScalarJob::predict(QSharedPointer<Session> session, QString target_feature,
+                                        Eigen::MatrixXd target_values, Direction direction) {
   // blocking call to predict scalars for given target particles
 
-  auto job = QSharedPointer<ShapeScalarJob>::create(session, target_feature, target_particles, JobType::Predict);
+  auto job = QSharedPointer<ShapeScalarJob>::create(session, target_feature, target_values, JobType::Predict);
   job->set_quiet_mode(true);
+  job->set_direction(direction);
 
   Eigen::VectorXd prediction;
 
@@ -104,8 +119,15 @@ void ShapeScalarJob::run_fit() {
   py::module np = py::module::import("numpy");
   py::module sw = py::module::import("shapeworks");
 
-  py::object A = np.attr("array")(all_particles_);
-  py::object B = np.attr("array")(all_scalars_);
+  py::object A;
+  py::object B;
+  if (direction_ == Direction::To_Scalar) {
+    A = np.attr("array")(all_particles_);
+    B = np.attr("array")(all_scalars_);
+  } else {
+    A = np.attr("array")(all_scalars_);
+    B = np.attr("array")(all_particles_);
+  }
 
   // returns a tuple of (png_raw_bytes, y_pred, mse)
   using ResultType = std::tuple<py::array, Eigen::MatrixXd, double>;
@@ -131,12 +153,13 @@ void ShapeScalarJob::run_prediction() {
   py::module sw = py::module::import("shapeworks");
 
   py::object does_mbpls_model_exist = sw.attr("shape_scalars").attr("does_mbpls_model_exist");
-  if (!does_mbpls_model_exist().cast<bool>()) {
+  if (needs_clear_ == true || !does_mbpls_model_exist().cast<bool>()) {
     SW_LOG("No MBPLS model exists, running fit");
     run_fit();
+    needs_clear_ = false;
   }
 
-  py::object new_x = np.attr("array")(target_particles_.transpose());
+  py::object new_x = np.attr("array")(target_values_.transpose());
   py::object run_prediction = sw.attr("shape_scalars").attr("pred_from_mbpls");
 
   using ResultType = Eigen::VectorXd;
diff --git a/Studio/Job/ShapeScalarJob.h b/Studio/Job/ShapeScalarJob.h
@@ -13,6 +13,7 @@ class ShapeScalarJob : public Job {
   Q_OBJECT
  public:
   enum class JobType { Find_Components, MSE_Plot, Predict };
+  enum class Direction { To_Shape, To_Scalar };
 
   ShapeScalarJob(QSharedPointer<Session> session, QString target_feature, Eigen::MatrixXd target_particles,
                  JobType job_type);
@@ -30,12 +31,22 @@ class ShapeScalarJob : public Job {
   static Eigen::VectorXd predict_scalars(QSharedPointer<Session> session, QString target_feature,
                                          Eigen::MatrixXd target_particles);
 
+  static Eigen::VectorXd predict_shape(QSharedPointer<Session> session, QString target_feature,
+                                       Eigen::MatrixXd target_particles);
+
+  static void clear_model() { needs_clear_ = true; };
+
+  void set_direction(Direction direction) { direction_ = direction; }
+
  private:
   void prep_data();
 
   void run_fit();
   void run_prediction();
 
+  static Eigen::VectorXd predict(QSharedPointer<Session> session, QString target_feature,
+                                 Eigen::MatrixXd target_particles, Direction direction);
+
   QSharedPointer<Session> session_;
 
   ParticleShapeStatistics stats_;
@@ -47,13 +58,16 @@ class ShapeScalarJob : public Job {
   Eigen::MatrixXd all_particles_;
   Eigen::MatrixXd all_scalars_;
 
-  Eigen::MatrixXd target_particles_;
+  Eigen::MatrixXd target_values_;
   Eigen::VectorXd prediction_;
 
   bool num_components_ = 3;
   int num_folds_ = 5;
   int max_components_ = 20;
 
+  Direction direction_{Direction::To_Scalar};
   JobType job_type_;
+
+  static std::atomic<bool> needs_clear_;
 };
 }  // namespace shapeworks
