Thanks to its simplicity, our method can be easily implemented on top ![]() Performs better than state-of-the-art on traditional single-label FGVC problemĪs well. Show that our method achieves superior performance in the new FGVC setting, and Label predictions, which in turn helps with better disentanglement. Ones, and (ii) allow finer-grained features to participate in coarser-grained Solution to our new problem, where we (i) leverage level-specificĬlassification heads to disentangle coarse-level features with fine-grained Thisĭiscovery enables us to design a very simple albeit surprisingly effective We then discover the key intuition that:Ĭoarse-level label prediction exacerbates fine-grained feature learning, yetįine-level feature betters the learning of coarse-level classifier. That most participants prefer multi-granularity labels, regardless whether theyĬonsider themselves experts. This new problem, we first conduct a comprehensive human study where we confirm "bird"->"Phoenicopteriformes"->"Phoenicopteridae"->"flamingo". ![]() Label hierarchy - so that our answer becomes That, we re-envisage the traditional setting of FGVC, from single-labelĬlassification, to that of top-down traversal of a pre-defined coarse-to-fine The real question is therefore - how can we tailor forĭifferent fine-grained definitions under divergent levels of expertise. To arrive at the former, for the majority of us non-experts just "bird" would While fine-grained visual classification (FGVC) strives ![]() Download a PDF of the paper titled Your "Flamingo" is My "Bird": Fine-Grained, or Not, by Dongliang Chang and 5 other authors Download PDF Abstract: Whether what you see in Figure 1 is a "flamingo" or a "bird", is the question
0 Comments
Leave a Reply. |