Hello, I have a doubt about the implementation of distanceBB_Gaussian
def distanceBB_Gaussian(box1, box2, sigma=1): off1 = np.array([ box1.center[0], box1.center[2], Quaternion(matrix=box1.rotation_matrix).degrees ]) off2 = np.array([ box2.center[0], box2.center[2], Quaternion(matrix=box2.rotation_matrix).degrees ]) dist = np.linalg.norm(off1 - off2) score = np.exp(-0.5 * dist / (sigma * sigma)) return score
Why do you use box1.center[0] and box1.center[2] that is x and z?
Why not use box1.center[0] and box1.center[1] that is x and y?
Hello, I have a doubt about the implementation of distanceBB_Gaussian
def distanceBB_Gaussian(box1, box2, sigma=1): off1 = np.array([ box1.center[0], box1.center[2], Quaternion(matrix=box1.rotation_matrix).degrees ]) off2 = np.array([ box2.center[0], box2.center[2], Quaternion(matrix=box2.rotation_matrix).degrees ]) dist = np.linalg.norm(off1 - off2) score = np.exp(-0.5 * dist / (sigma * sigma)) return score
Why do you use box1.center[0] and box1.center[2] that is x and z? Why not use box1.center[0] and box1.center[1] that is x and y?