Merge pull request #42 from philipqueen/main

Remove cv.sfm dependency

README.md

@@ -5,10 +5,6 @@
 ## Dependencies
 Install the pseyepy python library: [https://github.com/bensondaled/pseyepy](https://github.com/bensondaled/pseyepy)
 
-This project requires the sfm (structure from motion) OpenCV module, which requires you to compile OpenCV from source. This is a bit of a pain, but these links should help you get started: [SFM dependencies](https://docs.opencv.org/4.x/db/db8/tutorial_sfm_installation.html) [OpenCV module installation guide](https://github.com/opencv/opencv_contrib/blob/master/README.md)
-
-`cv.sfm` is only used 3 times in the codebase for the following functions: `fundamentalFromProjections`, `essentialFromFundamental`, `motionFromEssential`. So really, those functions should just be reimplemented in Python so the sfm module isn't needed. [Issue](https://github.com/jyjblrd/Mocap-Drones/issues/4).
-
 install npm and yarn
 
 ## Runing the code

computer_code/api/helpers.py

@@ -1,3 +1,4 @@
+from typing import List, Tuple
 import numpy as np
 from scipy import linalg, optimize, signal
 import cv2 as cv
@@ -240,6 +241,100 @@ def calculate_reprojection_error(image_points, object_point, camera_poses):
 
     return errors.mean()
 
+"""
+Original license for opencv sfm functions (applies to essential_from_fundamental, fundamental_from_projections, and motion_from_essential):
+Software License Agreement (BSD License)
+
+Copyright (c) 2009, Willow Garage, Inc.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions
+are met:
+ * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following
+   disclaimer in the documentation and/or other materials provided
+   with the distribution.
+ * Neither the name of Willow Garage, Inc. nor the names of its
+   contributors may be used to endorse or promote products derived
+   from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
+ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+POSSIBILITY OF SUCH DAMAGE.
+"""
+
+def essential_from_fundamental(F: np.ndarray, K1: np.ndarray, K2: np.ndarray) -> np.ndarray:
+    """
+    Calculate the essential matrix from the fundamental matrix (F) and camera matrices (K1, K2).
+
+    Adapted and modified from OpenCV's essentialFromFundamental function in the opencv_sfm module
+    """
+
+    assert F.shape == (3, 3), "F must be a 3x3 matrix"
+    assert K1.shape == (3, 3), "K1 must be a 3x3 matrix"
+    assert K2.shape == (3, 3), "K2 must be a 3x3 matrix"
+
+    E = np.dot(np.dot(K2.T, F), K1)
+    return E
+
+def fundamental_from_projections(P1: np.ndarray, P2: np.ndarray) -> np.ndarray:
+    """
+    Calculate the fundamental matrix from the projection matrices (P1, P2).
+    
+    Adapted and modified from OpenCV's fundamentalFromProjections function in the opencv_sfm module
+    """
+
+    assert P1.shape == (3, 4), "P1 must be a 3x4 matrix"
+    assert P2.shape == (3, 4), "P2 must be a 3x4 matrix"
+
+    F = np.zeros((3, 3))
+
+    X = np.array([
+        np.vstack((P1[1, :], P1[2, :])),
+        np.vstack((P1[2, :], P1[0, :])),
+        np.vstack((P1[0, :], P1[1, :]))
+    ])
+
+    Y = np.array([
+        np.vstack((P2[1, :], P2[2, :])),
+        np.vstack((P2[2, :], P2[0, :])),
+        np.vstack((P2[0, :], P2[1, :]))
+    ])
+
+    for i in range(3):
+        for j in range(3):
+            XY = np.vstack((X[j], Y[i]))
+            F[i, j] = np.linalg.det(XY)
+
+    return F
+
+def motion_from_essential(E: np.ndarray) -> Tuple[List[np.ndarray], List[np.ndarray]]:
+    """
+    Calculate the possible rotations and translations from the essential matrix (E).
+
+    Adapted and modified from OpenCV's motionFromEssential function in the opencv_sfm module
+    """
+    assert E.shape == (3, 3), "Essential matrix must be 3x3."
+
+    _, R1, R2, t = cv.decomposeEssentialMat(E)
+
+    rotations_matrices = [R1, R1, R2, R2]
+    translations = [t, -t, t, -t]
+
+    return rotations_matrices, translations
+
 
 def bundle_adjustment(image_points, camera_poses, socketio):
     cameras = Cameras.instance()
@@ -359,7 +454,7 @@ def find_point_correspondance_and_object_points(image_points, camera_poses, fram
     for i in range(1, len(camera_poses)):
         epipolar_lines = []
         for root_image_point in root_image_points:
-            F = cv.sfm.fundamentalFromProjections(Ps[root_image_point["camera"]], Ps[i])
+            F = fundamental_from_projections(Ps[root_image_point["camera"]], Ps[i])
             line = cv.computeCorrespondEpilines(np.array([root_image_point["point"]], dtype=np.float32), 1, F)
             epipolar_lines.append(line[0,0].tolist())
             frames[i] = drawlines(frames[i], line[0])

computer_code/api/index.py

@@ -1,4 +1,4 @@
-from helpers import camera_pose_to_serializable, calculate_reprojection_errors, bundle_adjustment, Cameras, triangulate_points
+from helpers import camera_pose_to_serializable, calculate_reprojection_errors, bundle_adjustment, Cameras, triangulate_points, essential_from_fundamental, motion_from_essential
 from KalmanFilter import KalmanFilter
 
 from flask import Flask, Response, request
@@ -244,8 +244,8 @@ def calculate_camera_pose(data):
         camera2_image_points = np.take(camera2_image_points, not_none_indicies, axis=0).astype(np.float32)
 
         F, _ = cv.findFundamentalMat(camera1_image_points, camera2_image_points, cv.FM_RANSAC, 1, 0.99999)
-        E = cv.sfm.essentialFromFundamental(F, cameras.get_camera_params(0)["intrinsic_matrix"], cameras.get_camera_params(1)["intrinsic_matrix"])
-        possible_Rs, possible_ts = cv.sfm.motionFromEssential(E)
+        E = essential_from_fundamental(F, cameras.get_camera_params(0)["intrinsic_matrix"], cameras.get_camera_params(1)["intrinsic_matrix"])
+        possible_Rs, possible_ts = motion_from_essential(E)
 
         R = None
         t = None