Abstract:
Learning image similarity underpins many vision tasks. Discriminative metric learning seeks a class embedding. The biggest challenge is learning a metric that generalizes from training to novel, but related, test samples. It should transfer to other object classes. The discriminative paradigm misses what complementary information? We need characteristics that separate classes and likely occur in novel categories, which is indicated if they are shared across training classes. This study investigates learning such characteristics without annotations or training data. Our novel triplet sampling strategy can be applied to recent ranking loss frameworks. Our approach improves deep metric learning performance regardless of network architecture and ranking loss, yielding state-of-the-art results on standard benchmark datasets.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
