〔Abstract〕 Objective To analyze the infl uence of the defl ection angle between the frontal facial plane and the camera optical axis on evaluation of facial paralysis. Methods A facial feature detection system using opencv+dlib based on computer vision was developed, then, a measurement model based on facial features for facial paralysis evaluation was constructed, and the infl uence of the defl ection angle between the frontal facial plane and the camera optical axis was analyzed. Results The experimental results demonstrate that the developed can achieve an average deviation within 1 mm between the real distance and the detection distance by computer vision when the angle between the frontal facial plane and the camera optical axis belongs to〔–30 °, 30 °〕, while the standard deviation of the gap of eye open in the horizontal and vertical direction belong to (1.13, 1.94) and (0.85, 1.24), respectively. Meanwhile, the standard deviation of the distance between the mouth corners and the nose tip belongs to (1.69, 2.78). The results present the developed system can obtain good detection accuracy within a set defl ection angle. Conclusion Geometric deviation will be generated on facial paralysis patients, and the developed facial detection system can recognize facial geometric features and achieve good detection accuracy. It provides a solution to perform evaluation of facial paralysis even the head of the subject rotate an angle within 30°.