CFC-DAOD is a domain adaptation extension to the Caltech Fish Counting domain generalization benchmark.
The goal is cross-domain object detection of a single class, "fish", in sonar videos. The source domain consists of data from one river, Kenai, and the target domain consists of data from another out-of-domain river, Channel. CFC-DAOD introduces new data from the target domain to be used for unsupervised domain adaptive object detection: 168k bounding box annotations in 29k frames sampled from 150 new videos captured over two days from 3 different sensors on the Channel river.
Paper | Code | Results | Date | Stars |
---|