Re-Ranking for Writer Identification and Writer Retrieval

摘要: Automatic writer identification is a common problem in document analysis. State-of-the-art methods typically focus on the feature extraction step with traditional or deep-learning-based techniques. In retrieval problems, re-ranking commonly used technique to improve results. Re-ranking refines an initial ranking result by using knowledge contained ranked result, e. g., exploiting nearest neighbor relations. To best of our knowledge, has not been for identification/retrieval. A possible reason might be that publicly available benchmark datasets contain only few samples per which makes less promising. We show based k-reciprocal relationships advantageous identification, even if are available. use these reciprocal two ways: encode them into new vectors, as originally proposed, integrate terms query-expansion. both techniques outperform baseline results mAP three datasets.