Domain Adaptive Neural Networks for Object Recognition

About

We propose a simple neural network model to deal with the domain adaptation problem in object recognition. Our model incorporates the Maximum Mean Discrepancy (MMD) measure as a regularization in the supervised learning to reduce the distribution mismatch between the source and target domains in the latent space. From experiments, we demonstrate that the MMD regularization is an effective tool to provide good domain adaptation models on both SURF features and raw image pixels of a particular image data set. We also show that our proposed model, preceded by the denoising auto-encoder pretraining, achieves better performance than recent benchmark models on the same data sets. This work represents the first study of MMD measure in the context of neural networks.

Muhammad Ghifary, W. Bastiaan Kleijn, Mengjie Zhang · 2014

Related benchmarks

Task	Dataset	Result
Unsupervised Domain Adaptation	Office-Home (test)	Average Accuracy57.6	347
Domain Adaptation	Office-31	Average Accuracy82.2	187
Image Classification	OFFICE Amazon → Webcam (test)	Accuracy53.6	8
Image Classification	OFFICE DSLR → Webcam (test)	Accuracy71.2	8
Multi-class classification	Office standard evaluation	Accuracy (A->W)0.536	7
Image Classification	OFFICE Webcam → DSLR (test)	Accuracy83.5	7
Multi-class classification	Office unsupervised standard	Average Accuracy59.9	6