Scene Recognition by Jointly Modeling Latent Topics
- In this project, the latent topic model is used to cluster visual words under different semantically
meaningful "topics" in an unsupervised manner. The histogram of latent topics is then used as input
to train/test scene classifiers.
- Such an approach is demonstrated to give superior accuracy than a number of traditional scene
recognition methods, including the basic Bag-of-Words (BoW) model.
- In our experiment, we 1) extract visual words from the regular dense grid of each image, 2)
infer the latent topics for these visual words using our method. The latent topics are visualized
in Figure 1.
- As we can see, the latent topics correspond very well with the underlying objects
and image regions. Without human annotation, our method can automatically cluster visual words
into different semantically meaningful groups.
- For experimental results on scene recognition performance, please read our paper.
- Latent Topic Discovering using LDA
- Scene Recognition by Jointly Modeling Latent Topics (Coming Soon...)