Merged in fix/snavelyPath (pull request #168)

Fixed AdaptAutoDiff

gtsam/geometry/SO3.cpp

@@ -32,18 +32,23 @@ SO3 SO3::Rodrigues(const Vector3& axis, double theta) {
   // get components of axis \omega
   double wx = axis(0), wy = axis(1), wz = axis(2);
 
-  double c = cos(theta), s = sin(theta), c_1 = 1 - c;
+  double costheta = cos(theta), sintheta = sin(theta), c_1 = 1 - costheta;
-  double wwTxx = wx * wx, wwTyy = wy * wy, wwTzz = wz * wz;
+  double wx_sintheta = wx * sintheta, wy_sintheta = wy * sintheta, wz_sintheta = wz * sintheta;
-  double swx = wx * s, swy = wy * s, swz = wz * s;
 
-  double C00 = c_1 * wwTxx, C01 = c_1 * wx * wy, C02 = c_1 * wx * wz;
+  double C00 = c_1 * wx * wx, C01 = c_1 * wx * wy, C02 = c_1 * wx * wz;
-  double C11 = c_1 * wwTyy, C12 = c_1 * wy * wz;
+  double C11 = c_1 * wy * wy, C12 = c_1 * wy * wz;
-  double C22 = c_1 * wwTzz;
+  double C22 = c_1 * wz * wz;
 
   Matrix3 R;
-  R << c + C00, -swz + C01, swy + C02, //
+  R(0, 0) =     costheta + C00;
-  swz + C01, c + C11, -swx + C12, //
+  R(1, 0) =  wz_sintheta + C01;
-  -swy + C02, swx + C12, c + C22;
+  R(2, 0) = -wy_sintheta + C02;
+  R(0, 1) = -wz_sintheta + C01;
+  R(1, 1) =     costheta + C11;
+  R(2, 1) =  wx_sintheta + C12;
+  R(0, 2) =  wy_sintheta + C02;
+  R(1, 2) = -wx_sintheta + C12;
+  R(2, 2) =     costheta + C22;
 
   return R;
 }

gtsam/nonlinear/AdaptAutoDiff.h

@@ -27,97 +27,46 @@
 
 namespace gtsam {
 
-namespace detail {
-
-// By default, we assume an Identity element
-template<typename T, typename structure_category>
-struct Origin { T operator()() { return traits<T>::Identity();} };
-
-// but simple manifolds don't have one, so we just use the default constructor
-template<typename T>
-struct Origin<T, manifold_tag> { T operator()() { return T();} };
-
-} // \ detail
-
-/**
- * Canonical is a template that creates canonical coordinates for a given type.
- * A simple manifold type (i.e., not a Lie Group) has no concept of identity,
- * hence in that case we use the value given by the default constructor T() as
- * the origin of a "canonical coordinate" parameterization.
- */
-template<typename T>
-struct Canonical {
-
-  GTSAM_CONCEPT_MANIFOLD_TYPE(T)
-
-  typedef traits<T> Traits;
-  enum { dimension = Traits::dimension };
-  typedef typename Traits::TangentVector TangentVector;
-  typedef typename Traits::structure_category Category;
-  typedef detail::Origin<T, Category> Origin;
-
-  static TangentVector Local(const T& other) {
-    return Traits::Local(Origin()(), other);
-  }
-
-  static T Retract(const TangentVector& v) {
-    return Traits::Retract(Origin()(), v);
-  }
-};
-
 /**
  * The AdaptAutoDiff class uses ceres-style autodiff to adapt a ceres-style
- * Function evaluation, i.e., a function F that defines an operator
+ * Function evaluation, i.e., a function FUNCTOR that defines an operator
- *   template<typename T> bool operator()(const T* const, const T* const, T* predicted) const;
+ *   template<typename T> bool operator()(const T* const, const T* const, T*
+ * predicted) const;
  * For now only binary operators are supported.
  */
-template<typename F, typename T, typename A1, typename A2>
+template <typename FUNCTOR, int M, int N1, int N2>
 class AdaptAutoDiff {
+  typedef Eigen::Matrix<double, M, N1, Eigen::RowMajor> RowMajor1;
+  typedef Eigen::Matrix<double, M, N2, Eigen::RowMajor> RowMajor2;
 
-  static const int N = traits<T>::dimension;
+  typedef Eigen::Matrix<double, M, 1> VectorT;
-  static const int M1 = traits<A1>::dimension;
+  typedef Eigen::Matrix<double, N1, 1> Vector1;
-  static const int M2 = traits<A2>::dimension;
+  typedef Eigen::Matrix<double, N2, 1> Vector2;
 
-  typedef Eigen::Matrix<double, N, M1, Eigen::RowMajor> RowMajor1;
+  FUNCTOR f;
-  typedef Eigen::Matrix<double, N, M2, Eigen::RowMajor> RowMajor2;
-
-  typedef Canonical<T> CanonicalT;
-  typedef Canonical<A1> Canonical1;
-  typedef Canonical<A2> Canonical2;
-
-  typedef typename CanonicalT::TangentVector VectorT;
-  typedef typename Canonical1::TangentVector Vector1;
-  typedef typename Canonical2::TangentVector Vector2;
-
-  F f;
-
-public:
-
-  T operator()(const A1& a1, const A2& a2, OptionalJacobian<N, M1> H1 = boost::none,
-      OptionalJacobian<N, M2> H2 = boost::none) {
 
+ public:
+  VectorT operator()(const Vector1& v1, const Vector2& v2,
+                     OptionalJacobian<M, N1> H1 = boost::none,
+                     OptionalJacobian<M, N2> H2 = boost::none) {
     using ceres::internal::AutoDiff;
 
-    // Make arguments
-    Vector1 v1 = Canonical1::Local(a1);
-    Vector2 v2 = Canonical2::Local(a2);
-
     bool success;
     VectorT result;
 
     if (H1 || H2) {
-
       // Get derivatives with AutoDiff
-      double *parameters[] = { v1.data(), v2.data() };
+      const double* parameters[] = {v1.data(), v2.data()};
-      double rowMajor1[N * M1], rowMajor2[N * M2]; // on the stack
+      double rowMajor1[M * N1], rowMajor2[M * N2];  // on the stack
-      double *jacobians[] = { rowMajor1, rowMajor2 };
+      double* jacobians[] = {rowMajor1, rowMajor2};
-      success = AutoDiff<F, double, 9, 3>::Differentiate(f, parameters, 2,
+      success = AutoDiff<FUNCTOR, double, N1, N2>::Differentiate(
-          result.data(), jacobians);
+          f, parameters, M, result.data(), jacobians);
 
       // Convert from row-major to columnn-major
-      // TODO: if this is a bottleneck (probably not!) fix Autodiff to be Column-Major
+      // TODO: if this is a bottleneck (probably not!) fix Autodiff to be
-      *H1 = Eigen::Map<RowMajor1>(rowMajor1);
+      // Column-Major
-      *H2 = Eigen::Map<RowMajor2>(rowMajor2);
+      if (H1) *H1 = Eigen::Map<RowMajor1>(rowMajor1);
+      if (H2) *H2 = Eigen::Map<RowMajor2>(rowMajor2);
 
     } else {
       // Apply the mapping, to get result
@@ -126,9 +75,8 @@ public:
     if (!success)
       throw std::runtime_error(
           "AdaptAutoDiff: function call resulted in failure");
-    return CanonicalT::Retract(result);
+    return result;
   }
-
 };
 
-}
+}  // namespace gtsam

gtsam/nonlinear/LevenbergMarquardtOptimizer.cpp

@@ -25,6 +25,9 @@
 
 #include <boost/algorithm/string.hpp>
 #include <boost/range/adaptor/map.hpp>
+#include <boost/format.hpp>
+#include <boost/timer/timer.hpp>
+
 #include <fstream>
 #include <limits>
 #include <string>
@@ -236,7 +239,7 @@ void LevenbergMarquardtOptimizer::iterate() {
 
   // Keep increasing lambda until we make make progress
   while (true) {
-
+    boost::timer::cpu_timer timer;
     if (lmVerbosity >= LevenbergMarquardtParams::TRYLAMBDA)
       cout << "trying lambda = " << state_.lambda << endl;
 
@@ -248,7 +251,7 @@ void LevenbergMarquardtOptimizer::iterate() {
     double modelFidelity = 0.0;
     bool step_is_successful = false;
     bool stopSearchingLambda = false;
-    double newError = numeric_limits<double>::infinity();
+    double newError = numeric_limits<double>::infinity(), costChange;
     Values newValues;
     VectorValues delta;
 
@@ -261,8 +264,6 @@ void LevenbergMarquardtOptimizer::iterate() {
       systemSolvedSuccessfully = false;
     }
 
-    double linearizedCostChange = 0,
-           newlinearizedError = 0;
     if (systemSolvedSuccessfully) {
       state_.reuseDiagonal = true;
 
@@ -272,9 +273,9 @@ void LevenbergMarquardtOptimizer::iterate() {
         delta.print("delta");
 
       // cost change in the linearized system (old - new)
-      newlinearizedError = linear->error(delta);
+      double newlinearizedError = linear->error(delta);
 
-      linearizedCostChange = state_.error - newlinearizedError;
+      double linearizedCostChange = state_.error - newlinearizedError;
       if (lmVerbosity >= LevenbergMarquardtParams::TRYLAMBDA)
               cout << "newlinearizedError = " << newlinearizedError <<
               "  linearizedCostChange = " << linearizedCostChange << endl;
@@ -299,7 +300,7 @@ void LevenbergMarquardtOptimizer::iterate() {
               << ") new (tentative) error (" << newError << ")" << endl;
 
         // cost change in the original, nonlinear system (old - new)
-        double costChange = state_.error - newError;
+        costChange = state_.error - newError;
 
         if (linearizedCostChange > 1e-20) { // the (linear) error has to decrease to satisfy this condition
           // fidelity of linearized model VS original system between
@@ -322,9 +323,13 @@ void LevenbergMarquardtOptimizer::iterate() {
     }
 
     if (lmVerbosity == LevenbergMarquardtParams::SUMMARY) {
-      cout << "[" << state_.iterations << "]: " << "new error = " << newlinearizedError
+      // do timing
-           << ", delta = " << linearizedCostChange << ", lambda = " << state_.lambda
+      double iterationTime = 1e-9 * timer.elapsed().wall;
-           << ", success = " << systemSolvedSuccessfully << std::endl;
+      if (state_.iterations == 0)
+        cout << "iter      cost      cost_change    lambda  success iter_time" << endl;
+      cout << boost::format("% 4d % 8e   % 3.2e   % 3.2e  % 4d   % 3.2e") %
+                  state_.iterations % newError % costChange % state_.lambda %
+                  systemSolvedSuccessfully % iterationTime << endl;
     }
 
     ++state_.totalNumberInnerIterations;

gtsam/nonlinear/tests/testAdaptAutoDiff.cpp

@@ -36,81 +36,43 @@ using boost::assign::map_list_of;
 namespace gtsam {
 
 // Special version of Cal3Bundler so that default constructor = 0,0,0
-struct Cal: public Cal3Bundler {
+struct Cal3Bundler0 : public Cal3Bundler {
-  Cal(double f = 0, double k1 = 0, double k2 = 0, double u0 = 0, double v0 = 0) :
+  Cal3Bundler0(double f = 0, double k1 = 0, double k2 = 0, double u0 = 0,
-      Cal3Bundler(f, k1, k2, u0, v0) {
+               double v0 = 0)
+      : Cal3Bundler(f, k1, k2, u0, v0) {}
+  Cal3Bundler0 retract(const Vector& d) const {
+    return Cal3Bundler0(fx() + d(0), k1() + d(1), k2() + d(2), u0(), v0());
   }
-  Cal retract(const Vector& d) const {
+  Vector3 localCoordinates(const Cal3Bundler0& T2) const {
-    return Cal(fx() + d(0), k1() + d(1), k2() + d(2), u0(), v0());
-  }
-  Vector3 localCoordinates(const Cal& T2) const {
     return T2.vector() - vector();
   }
 };
 
-template<>
+template <>
-struct traits<Cal> : public internal::Manifold<Cal> {};
+struct traits<Cal3Bundler0> : public internal::Manifold<Cal3Bundler0> {};
 
 // With that, camera below behaves like Snavely's 9-dim vector
-typedef PinholeCamera<Cal> Camera;
+typedef PinholeCamera<Cal3Bundler0> Camera;
-
 }
 
 using namespace std;
 using namespace gtsam;
 
 /* ************************************************************************* */
-// charts
+// Check that ceres rotation convention is the same
-TEST(AdaptAutoDiff, Canonical) {
+TEST(AdaptAutoDiff, Rotation) {
-
+  Vector3 axisAngle(0.1, 0.2, 0.3);
-  Canonical<Point2> chart1;
+  Matrix3 expected = Rot3::rodriguez(axisAngle).matrix();
-  EXPECT(chart1.Local(Point2(1, 0))==Vector2(1, 0));
+  Matrix3 actual;
-  EXPECT(chart1.Retract(Vector2(1, 0))==Point2(1, 0));
+  ceres::AngleAxisToRotationMatrix(axisAngle.data(), actual.data());
-
+  EXPECT(assert_equal(expected, actual));
-  Vector v2(2);
-  v2 << 1, 0;
-  Canonical<Vector2> chart2;
-  EXPECT(assert_equal(v2, chart2.Local(Vector2(1, 0))));
-  EXPECT(chart2.Retract(v2)==Vector2(1, 0));
-
-  Canonical<double> chart3;
-  Eigen::Matrix<double, 1, 1> v1;
-  v1 << 1;
-  EXPECT(chart3.Local(1)==v1);
-  EXPECT_DOUBLES_EQUAL(chart3.Retract(v1), 1, 1e-9);
-
-  Canonical<Point3> chart4;
-  Point3 point(1, 2, 3);
-  Vector v3(3);
-  v3 << 1, 2, 3;
-  EXPECT(assert_equal(v3, chart4.Local(point)));
-  EXPECT(assert_equal(chart4.Retract(v3), point));
-
-  Canonical<Pose3> chart5;
-  Pose3 pose(Rot3::identity(), point);
-  Vector v6(6);
-  v6 << 0, 0, 0, 1, 2, 3;
-  EXPECT(assert_equal(v6, chart5.Local(pose)));
-  EXPECT(assert_equal(chart5.Retract(v6), pose));
-
-  Canonical<Cal> chart6;
-  Cal cal0;
-  Vector z3 = Vector3::Zero();
-  EXPECT(assert_equal(z3, chart6.Local(cal0)));
-  EXPECT(assert_equal(chart6.Retract(z3), cal0));
-
-  Canonical<Camera> chart7;
-  Camera camera(Pose3(), cal0);
-  Vector z9 = Vector9::Zero();
-  EXPECT(assert_equal(z9, chart7.Local(camera)));
-  EXPECT(assert_equal(chart7.Retract(z9), camera));
 }
 
 /* ************************************************************************* */
 // Some Ceres Snippets copied for testing
 // Copyright 2010, 2011, 2012 Google Inc. All rights reserved.
-template<typename T> inline T &RowMajorAccess(T *base, int rows, int cols,
+template <typename T>
-    int i, int j) {
+inline T& RowMajorAccess(T* base, int rows, int cols, int i, int j) {
   return base[cols * i + j];
 }
 
@@ -124,14 +86,14 @@ inline double RandDouble() {
 struct Projective {
   // Function that takes P and X as separate vectors:
   //   P, X -> x
-  template<typename A>
+  template <typename A>
   bool operator()(A const P[12], A const X[4], A x[2]) const {
     A PX[3];
     for (int i = 0; i < 3; ++i) {
-      PX[i] = RowMajorAccess(P, 3, 4, i, 0) * X[0]
+      PX[i] = RowMajorAccess(P, 3, 4, i, 0) * X[0] +
-          + RowMajorAccess(P, 3, 4, i, 1) * X[1]
+              RowMajorAccess(P, 3, 4, i, 1) * X[1] +
-          + RowMajorAccess(P, 3, 4, i, 2) * X[2]
+              RowMajorAccess(P, 3, 4, i, 2) * X[2] +
-          + RowMajorAccess(P, 3, 4, i, 3) * X[3];
+              RowMajorAccess(P, 3, 4, i, 3) * X[3];
     }
     if (PX[2] != 0.0) {
       x[0] = PX[0] / PX[2];
@@ -160,29 +122,31 @@ TEST(AdaptAutoDiff, AutoDiff) {
   Projective projective;
 
   // Make arguments
-  typedef Eigen::Matrix<double, 3, 4, Eigen::RowMajor> M;
+  typedef Eigen::Matrix<double, 3, 4, Eigen::RowMajor> RowMajorMatrix34;
-  M P;
+  RowMajorMatrix34 P;
   P << 1, 0, 0, 0, 0, 1, 0, 5, 0, 0, 1, 0;
   Vector4 X(10, 0, 5, 1);
 
   // Apply the mapping, to get image point b_x.
   Vector expected = Vector2(2, 1);
   Vector2 actual = projective(P, X);
-  EXPECT(assert_equal(expected,actual,1e-9));
+  EXPECT(assert_equal(expected, actual, 1e-9));
 
   // Get expected derivatives
-  Matrix E1 = numericalDerivative21<Vector2, M, Vector4>(Projective(), P, X);
+  Matrix E1 = numericalDerivative21<Vector2, RowMajorMatrix34, Vector4>(
-  Matrix E2 = numericalDerivative22<Vector2, M, Vector4>(Projective(), P, X);
+      Projective(), P, X);
+  Matrix E2 = numericalDerivative22<Vector2, RowMajorMatrix34, Vector4>(
+      Projective(), P, X);
 
   // Get derivatives with AutoDiff
   Vector2 actual2;
   MatrixRowMajor H1(2, 12), H2(2, 4);
-  double *parameters[] = { P.data(), X.data() };
+  double* parameters[] = {P.data(), X.data()};
-  double *jacobians[] = { H1.data(), H2.data() };
+  double* jacobians[] = {H1.data(), H2.data()};
-  CHECK(
+  CHECK((AutoDiff<Projective, double, 12, 4>::Differentiate(
-      (AutoDiff<Projective, double, 12, 4>::Differentiate( projective, parameters, 2, actual2.data(), jacobians)));
+      projective, parameters, 2, actual2.data(), jacobians)));
-  EXPECT(assert_equal(E1,H1,1e-8));
+  EXPECT(assert_equal(E1, H1, 1e-8));
-  EXPECT(assert_equal(E2,H2,1e-8));
+  EXPECT(assert_equal(E2, H2, 1e-8));
 }
 
 /* ************************************************************************* */
@@ -197,78 +161,85 @@ Vector2 adapted(const Vector9& P, const Vector3& X) {
     throw std::runtime_error("Snavely fail");
 }
 
+/* ************************************************************************* */
+namespace example {
+Camera camera(Pose3(Rot3().retract(Vector3(0.1, 0.2, 0.3)), Point3(0, 5, 0)),
+              Cal3Bundler0(1, 0, 0));
+Point3 point(10, 0, -5);  // negative Z-axis convention of Snavely!
+Vector9 P = Camera().localCoordinates(camera);
+Vector3 X = Point3::Logmap(point);
+Vector2 expectedMeasurement(1.2431567, 1.2525694);
+Matrix E1 = numericalDerivative21<Vector2, Vector9, Vector3>(adapted, P, X);
+Matrix E2 = numericalDerivative22<Vector2, Vector9, Vector3>(adapted, P, X);
+}
+
+/* ************************************************************************* */
+// Check that Local worked as expected
+TEST(AdaptAutoDiff, Local) {
+  using namespace example;
+  Vector9 expectedP = (Vector9() << 0.1, 0.2, 0.3, 0, 5, 0, 1, 0, 0).finished();
+  EXPECT(equal_with_abs_tol(expectedP, P));
+  Vector3 expectedX(10, 0, -5);  // negative Z-axis convention of Snavely!
+  EXPECT(equal_with_abs_tol(expectedX, X));
+}
+
+/* ************************************************************************* */
+// Test Ceres AutoDiff
 TEST(AdaptAutoDiff, AutoDiff2) {
+  using namespace example;
   using ceres::internal::AutoDiff;
 
+  // Apply the mapping, to get image point b_x.
+  Vector2 actual = adapted(P, X);
+  EXPECT(assert_equal(expectedMeasurement, actual, 1e-6));
+
   // Instantiate function
   SnavelyProjection snavely;
 
-  // Make arguments
-  Vector9 P; // zero rotation, (0,5,0) translation, focal length 1
-  P << 0, 0, 0, 0, 5, 0, 1, 0, 0;
-  Vector3 X(10, 0, -5); // negative Z-axis convention of Snavely!
-
-  // Apply the mapping, to get image point b_x.
-  Vector expected = Vector2(2, 1);
-  Vector2 actual = adapted(P, X);
-  EXPECT(assert_equal(expected,actual,1e-9));
-
-  // Get expected derivatives
-  Matrix E1 = numericalDerivative21<Vector2, Vector9, Vector3>(adapted, P, X);
-  Matrix E2 = numericalDerivative22<Vector2, Vector9, Vector3>(adapted, P, X);
-
   // Get derivatives with AutoDiff
   Vector2 actual2;
   MatrixRowMajor H1(2, 9), H2(2, 3);
-  double *parameters[] = { P.data(), X.data() };
+  double* parameters[] = {P.data(), X.data()};
-  double *jacobians[] = { H1.data(), H2.data() };
+  double* jacobians[] = {H1.data(), H2.data()};
-  CHECK(
+  CHECK((AutoDiff<SnavelyProjection, double, 9, 3>::Differentiate(
-      (AutoDiff<SnavelyProjection, double, 9, 3>::Differentiate( snavely, parameters, 2, actual2.data(), jacobians)));
+      snavely, parameters, 2, actual2.data(), jacobians)));
-  EXPECT(assert_equal(E1,H1,1e-8));
+  EXPECT(assert_equal(E1, H1, 1e-8));
-  EXPECT(assert_equal(E2,H2,1e-8));
+  EXPECT(assert_equal(E2, H2, 1e-8));
 }
 
 /* ************************************************************************* */
 // Test AutoDiff wrapper Snavely
-TEST(AdaptAutoDiff, AutoDiff3) {
+TEST(AdaptAutoDiff, AdaptAutoDiff) {
+  using namespace example;
 
-  // Make arguments
+  typedef AdaptAutoDiff<SnavelyProjection, 2, 9, 3> Adaptor;
-  Camera P(Pose3(Rot3(), Point3(0, 5, 0)), Cal(1, 0, 0));
-  Point3 X(10, 0, -5); // negative Z-axis convention of Snavely!
-
-  typedef AdaptAutoDiff<SnavelyProjection, Point2, Camera, Point3> Adaptor;
   Adaptor snavely;
 
   // Apply the mapping, to get image point b_x.
-  Point2 expected(2, 1);
+  Vector2 actual = snavely(P, X);
-  Point2 actual = snavely(P, X);
+  EXPECT(assert_equal(expectedMeasurement, actual, 1e-6));
-  EXPECT(assert_equal(expected,actual,1e-9));
-
-//  // Get expected derivatives
-  Matrix E1 = numericalDerivative21<Point2, Camera, Point3>(Adaptor(), P, X);
-  Matrix E2 = numericalDerivative22<Point2, Camera, Point3>(Adaptor(), P, X);
 
   // Get derivatives with AutoDiff, not gives RowMajor results!
   Matrix29 H1;
   Matrix23 H2;
-  Point2 actual2 = snavely(P, X, H1, H2);
+  Vector2 actual2 = snavely(P, X, H1, H2);
-  EXPECT(assert_equal(expected,actual2,1e-9));
+  EXPECT(assert_equal(expectedMeasurement, actual2, 1e-6));
-  EXPECT(assert_equal(E1,H1,1e-8));
+  EXPECT(assert_equal(E1, H1, 1e-8));
-  EXPECT(assert_equal(E2,H2,1e-8));
+  EXPECT(assert_equal(E2, H2, 1e-8));
 }
 
 /* ************************************************************************* */
 // Test AutoDiff wrapper in an expression
-TEST(AdaptAutoDiff, Snavely) {
+TEST(AdaptAutoDiff, SnavelyExpression) {
-  Expression<Camera> P(1);
+  Expression<Vector9> P(1);
-  Expression<Point3> X(2);
+  Expression<Vector3> X(2);
-  typedef AdaptAutoDiff<SnavelyProjection, Point2, Camera, Point3> Adaptor;
+  typedef AdaptAutoDiff<SnavelyProjection, 2, 9, 3> Adaptor;
-  Expression<Point2> expression(Adaptor(), P, X);
+  Expression<Vector2> expression(Adaptor(), P, X);
+  EXPECT_LONGS_EQUAL(
+      sizeof(internal::BinaryExpression<Vector2, Vector9, Vector3>::Record),
+      expression.traceSize());
 #ifdef GTSAM_USE_QUATERNIONS
-  EXPECT_LONGS_EQUAL(384,expression.traceSize()); // Todo, should be zero
+  EXPECT_LONGS_EQUAL(336, expression.traceSize());
-#else
-  EXPECT_LONGS_EQUAL(sizeof(internal::BinaryExpression<Point2, Camera, Point3>::Record),
-                     expression.traceSize());  // Todo, should be zero
 #endif
   set<Key> expected = list_of(1)(2);
   EXPECT(expected == expression.keys());
@@ -280,4 +251,3 @@ int main() {
   return TestRegistry::runAllTests(tr);
 }
 /* ************************************************************************* */
-

timing/timeAdaptAutoDiff.cpp

@@ -57,16 +57,19 @@ int main() {
   f1 = boost::make_shared<GeneralSFMFactor<Camera, Point3> >(z, model, 1, 2);
   time("GeneralSFMFactor<Camera>                : ", f1, values);
 
-  // AdaptAutoDiff
-  typedef AdaptAutoDiff<SnavelyProjection, Point2, Camera, Point3> AdaptedSnavely;
-  Point2_ expression(AdaptedSnavely(), camera, point);
-  f2 = boost::make_shared<ExpressionFactor<Point2> >(model, z, expression);
-  time("Point2_(AdaptedSnavely(), camera, point): ", f2, values);
-
   // ExpressionFactor
   Point2_ expression2(camera, &Camera::project2, point);
   f3 = boost::make_shared<ExpressionFactor<Point2> >(model, z, expression2);
   time("Point2_(camera, &Camera::project, point): ", f3, values);
 
+  // AdaptAutoDiff
+  values.clear();
+  values.insert(1,Vector9(Vector9::Zero()));
+  values.insert(2,Vector3(0,0,1));
+  typedef AdaptAutoDiff<SnavelyProjection, 2, 9, 3> AdaptedSnavely;
+  Expression<Vector2> expression(AdaptedSnavely(), Expression<Vector9>(1), Expression<Vector3>(2));
+  f2 = boost::make_shared<ExpressionFactor<Vector2> >(model, z.vector(), expression);
+  time("Point2_(AdaptedSnavely(), camera, point): ", f2, values);
+
   return 0;
 }

timing/timeSFMBAL.cpp

@@ -34,7 +34,7 @@
 using namespace std;
 using namespace gtsam;
 
-#define USE_GTSAM_FACTOR
+//#define USE_GTSAM_FACTOR
 #ifdef USE_GTSAM_FACTOR
 #include <gtsam/slam/GeneralSFMFactor.h>
 #include <gtsam/geometry/Cal3Bundler.h>
@@ -45,6 +45,7 @@ typedef GeneralSFMFactor<Camera, Point3> SfmFactor;
 #include <gtsam/3rdparty/ceres/example.h>
 #include <gtsam/nonlinear/ExpressionFactor.h>
 #include <gtsam/nonlinear/AdaptAutoDiff.h>
+// See http://www.cs.cornell.edu/~snavely/bundler/bundler-v0.3-manual.html for conventions
 // Special version of Cal3Bundler so that default constructor = 0,0,0
 struct CeresCalibration: public Cal3Bundler {
   CeresCalibration(double f = 0, double k1 = 0, double k2 = 0, double u0 = 0,
@@ -76,14 +77,15 @@ int main(int argc, char* argv[]) {
   using symbol_shorthand::P;
 
   // Load BAL file (default is tiny)
-  string defaultFilename = findExampleDataFile("dubrovnik-3-7-pre");
+  //string defaultFilename = findExampleDataFile("dubrovnik-3-7-pre");
+  string defaultFilename = "/home/frank/problem-16-22106-pre.txt";
   SfM_data db;
   bool success = readBAL(argc > 1 ? argv[1] : defaultFilename, db);
   if (!success)
     throw runtime_error("Could not access file!");
 
 #ifndef USE_GTSAM_FACTOR
-  AdaptAutoDiff<SnavelyProjection, Point2, Camera, Point3> snavely;
+  AdaptAutoDiff<SnavelyProjection, 2, 9, 3> snavely;
 #endif
 
   // Build graph
@@ -92,14 +94,15 @@ int main(int argc, char* argv[]) {
   for (size_t j = 0; j < db.number_tracks(); j++) {
     BOOST_FOREACH (const SfM_Measurement& m, db.tracks[j].measurements) {
       size_t i = m.first;
-      Point2 measurement = m.second;
+      Point2 z = m.second;
 #ifdef USE_GTSAM_FACTOR
-      graph.push_back(SfmFactor(measurement, unit2, i, P(j)));
+      graph.push_back(SfmFactor(z, unit2, i, P(j)));
 #else
-      Expression<Camera> camera_(i);
+      Expression<Vector9> camera_(i);
-      Expression<Point3> point_(P(j));
+      Expression<Vector3> point_(P(j));
-      graph.addExpressionFactor(unit2, measurement,
+      // Snavely expects measurements in OpenGL format, with y increasing upwards
-          Expression<Point2>(snavely, camera_, point_));
+      graph.addExpressionFactor(unit2, Vector2(z.x(), -z.y()),
+                                Expression<Vector2>(snavely, camera_, point_));
 #endif
     }
   }
@@ -110,21 +113,33 @@ int main(int argc, char* argv[]) {
 #ifdef USE_GTSAM_FACTOR
     initial.insert((i++), camera);
 #else
-    Camera ceresCamera(camera.pose(), camera.calibration());
+    // readBAL converts to GTSAM format, so we need to convert back !
-    initial.insert((i++), ceresCamera);
+    Camera ceresCamera(gtsam2openGL(camera.pose()), camera.calibration());
+    Vector9 v9 = Camera().localCoordinates(ceresCamera);
+    initial.insert((i++), v9);
 #endif
   }
-  BOOST_FOREACH(const SfM_Track& track, db.tracks)
+  BOOST_FOREACH(const SfM_Track& track, db.tracks) {
+#ifdef USE_GTSAM_FACTOR
     initial.insert(P(j++), track.p);
+#else
+    Vector3 v3 = track.p.vector();
+    initial.insert(P(j++), v3);
+#endif
+  }
 
-  // Check projection
+  // Check projection of first point in first camera
   Point2 expected = db.tracks.front().measurements.front().second;
+#ifdef USE_GTSAM_FACTOR
   Camera camera = initial.at<Camera>(0);
   Point3 point = initial.at<Point3>(P(0));
-#ifdef USE_GTSAM_FACTOR
   Point2 actual = camera.project(point);
 #else
-  Point2 actual = snavely(camera, point);
+  Vector9 camera = initial.at<Vector9>(0);
+  Vector3 point = initial.at<Vector3>(P(0));
+  Point2 z = snavely(camera, point);
+  // Again: fix y to increase upwards
+  Point2 actual(z.x(), -z.y());
 #endif
   assert_equal(expected,actual,10);
 
@@ -146,8 +161,7 @@ int main(int argc, char* argv[]) {
   LevenbergMarquardtParams params;
   LevenbergMarquardtParams::SetCeresDefaults(&params);
   params.setOrdering(ordering);
-  params.setVerbosity("ERROR");
+  params.setVerbosityLM("SUMMARY");
-  params.setVerbosityLM("TRYLAMBDA");
   LevenbergMarquardtOptimizer lm(graph, initial, params);
   Values result = lm.optimize();