Computer vision has been successfully used in real-world recognition problems, where state-of-the-art recognition algorithms focus on training classifiers or regressors from large training sets. Feature extraction is critical for final performance, especially in the age of big data. While most prevalent features are hand-crafted (SIFT, HOG, LBP), they are not adaptive and rely heavily on supervised learning with costly labeled data.
In contrast, unlabeled data are cheap and abundant, especially from online photo and video sharing platforms such as Flickr, Picasa, Facebook, Google Images, and YouTube. This special issue aims to solicit recent state-of-the-art achievements on learning discriminative visual features from big data.
Important Dates:
All submissions must follow the regular full-length Pattern Recognition paper format. The submission website is: http://ees.elsevier.com/pr/. Please select “SI : DFL from Big data” as the Article Type.
The Call for Papers can be downloaded from here.