Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Face recognition model training method, storage medium and intelligent door lock
The invention discloses a face recognition model training method, a storage medium and an intelligent door lock, and the training method comprises the steps: obtaining a first training image set which comprises a sample image and a depth image corresponding to the sample image; inputting the sample image and the depth image into an initial face recognition model for training to obtain a sample feature vector, a depth feature vector and a fusion feature; determining a loss function of the initial face recognition model based on the sample feature vector, the depth feature vector and the fusion feature; and according to the loss function of the initial face recognition model, adjusting parameters of the initial face recognition model to obtain a target face recognition model. According to the training method, million-level color images and corresponding depth images do not need to be collected, the training cost is low, and the target face recognition model obtained through training of the training method has the advantages of being high in face recognition precision, low in face recognition energy consumption and high in face recognition speed.
本发明公开了一种人脸识别模型的训练方法、存储介质及智能门锁，该训练方法包括：获取第一训练图像集，第一训练图像集包括样本图像和与样本图像对应的深度图像；将样本图像和深度图像，输入至初始人脸识别模型进行训练，得到样本特征向量、深度特征向量、融合特征；基于样本特征向量、深度特征向量、融合特征确定初始人脸识别模型的损失函数；根据初始人脸识别模型的损失函数，调整初始人脸识别模型的参数，得到目标人脸识别模型。该训练方法无需采集百万级别的彩色图像及其对应的深度图像，具有训练成本低的优点，且该训练方法训练得到的目标人脸识别模型具有人脸识别精度高，人脸识别能耗低，人脸识别速度快的优点。
Face recognition model training method, storage medium and intelligent door lock
The invention discloses a face recognition model training method, a storage medium and an intelligent door lock, and the training method comprises the steps: obtaining a first training image set which comprises a sample image and a depth image corresponding to the sample image; inputting the sample image and the depth image into an initial face recognition model for training to obtain a sample feature vector, a depth feature vector and a fusion feature; determining a loss function of the initial face recognition model based on the sample feature vector, the depth feature vector and the fusion feature; and according to the loss function of the initial face recognition model, adjusting parameters of the initial face recognition model to obtain a target face recognition model. According to the training method, million-level color images and corresponding depth images do not need to be collected, the training cost is low, and the target face recognition model obtained through training of the training method has the advantages of being high in face recognition precision, low in face recognition energy consumption and high in face recognition speed.
本发明公开了一种人脸识别模型的训练方法、存储介质及智能门锁，该训练方法包括：获取第一训练图像集，第一训练图像集包括样本图像和与样本图像对应的深度图像；将样本图像和深度图像，输入至初始人脸识别模型进行训练，得到样本特征向量、深度特征向量、融合特征；基于样本特征向量、深度特征向量、融合特征确定初始人脸识别模型的损失函数；根据初始人脸识别模型的损失函数，调整初始人脸识别模型的参数，得到目标人脸识别模型。该训练方法无需采集百万级别的彩色图像及其对应的深度图像，具有训练成本低的优点，且该训练方法训练得到的目标人脸识别模型具有人脸识别精度高，人脸识别能耗低，人脸识别速度快的优点。
Face recognition model training method, storage medium and intelligent door lock
人脸识别模型的训练方法、存储介质及智能门锁
CAI ZHONGYIN (Autor:in)
01.11.2022
Patent
Elektronische Ressource
Chinesisch
Intelligent door lock based on face recognition function
| Europäisches Patentamt | 2023
Intelligent door lock control method, intelligent door lock system and storage medium
| Europäisches Patentamt | 2023
Non-embedded intelligent door lock based on face recognition
| Europäisches Patentamt | 2020