domain adaptation benchmark datasets

Citation Author(s):
Kaizhong
Jin
Submitted by:
Kaizhong Jin
Last updated:
Wed, 02/21/2024 - 00:04
DOI:
10.21227/fkye-zp36
License:
116 Views
Categories:
Keywords:
0
0 ratings - Please login to submit your rating.

Abstract 

We evaluate our approach on three popular domain adaptation benchmark datasets. The first one is Office-Caltech10 dataset, which contains images of 10 object categories from an office environment (e.g., keyboard, laptop) in 4 sources: Amazon, Caltech256, DSLR, and Webcam. We encode each source into 4096-dimensional feature vectors. Using each source as a domain, we get four domains leading to 12 domain adaptation tasks. The second one is Office-Home dataset, which contains images of 65 object categories found typically in Office and Home settings. The dataset includes 4 domains: Art, Clipart, Realworld and Product. We encode each domain into 4096-dimensional feature vectors. Similarly, twelve domain adaptation tasks are conducted by taking one sub-dataset as the source domain and the other one as the target domain. The third one is Amazon review dataset that is used for sentimental analysis of text. The dataset contains Amazon reviews on 4 domains: Book, DVD, Kitchen and Electronics, yielding 12 domain adaptation tasks of source-target domain pairs.