
Query – Free Clothing Retrieval via Implicit Relevance Feedback
Abstract
we investigate a new type of clothing retrieval task in which the target image resides only in the users’ mind, which is a more realistic scenario for real-world applications;
Conclusion and Future Work
In this Query – Free Clothing Retrieval via Implicit Relevance Feedback paper, we investigate a new form of clothing retrieval problem in which an image of the target item resides only in the user’s mind.
Based on heterogeneous features extracted from clothing attributes using deep CNNs, a significant advantage of our search-dependent re-weighting scheme is that it models the variability of human decision-making through implicit feedback.
Experimental results show that the proposed algorithm consistently outperforms previously developed algorithms based on a pre-defined image similarity metric.
It is interesting to see if theres a more unbiased setting for the mental image game. We will leave it as a future work.