Method, device, and storage medium for targeted adversarial discriminative domain adaptation

Abstract

Embodiments of the present disclosure provide a method, a device, and a storage medium for targeted adversarial discriminative domain adaptation (T-ADDA). The method includes pre-training a source model including a source feature encoder and a source classifier, adapting a target feature encoder, and generating a target model by concatenating the adapted target feature encoder with the pre-trained source classifier. Adapting the target feature encoder includes configuring the pre-trained source feature encoder to be an initial target feature encoder for generating target feature vectors in each target class; adjusting a domain discriminator according to an adversarial domain discrimination loss; adjusting the initial target feature encoder according to a generative adversarial network (GAN) loss; and further adjusting the initial target feature encoder to generate the target feature encoder according to a feature class matching loss using labeled target feature vectors and corresponding source feature class centers.