Combining Convolutional Neural Network and Markov Random Field for Semantic Image Retrieval

Figure 2

Semantic retrieval performance (MAPs % and P@10s %) over the comparative group: a difficult query set and an easy query set on two datasets.

(a) MAPs on MIR Flickr 2011

(b) P@10s on MIR Flickr 2011

(c) MAPs on NUS-WIDE

(d) P@10s on NUS-WIDE