The following visualizations correspond Figures 3, 4, and 5 from the paper. Data were collected in dives by the SeaBed AUV at Hannibal seamount in Panama in 2015. Move the time slider to explore images from the dataset, and observe the corresponding topic distribution or perplexity in the plots.

Figure 3 Results for three unsupervised topic models versus hand-annotated terrain labels in (b) for the Mission I dataset. Example images from the dataset are shown in (a). To generate plots (c,d,e), visual words are extracted from an image at time t and assigned a topic label z i as described in the text. The proportion of words in the image at time t assigned to each topic label is shown on the y-axis, where different topics are represented by colors.

The hybrid HDP-CAE model (c), using much fewer, more abstract features, is able to define richer topics that correspond more directly to useful visual phenomena then the purely neural model (d) or the HDP model using standard image features (e).

Figure 4 Results for three unsupervised topic models versus annotated labels in (b) for the Mission II dataset. Example images from the dataset are shown in (a). The hybrid HDP-CAE model (c) again outperforms the purely neural model (d) and the HDP model using standard image features (e). However, the hybrid model fails recognize some of the more transient topics, such as the crustacean swarm at (7).

Figure 5 Correlation of perplexity score of the three unsupervised models (c,d,e) with annotated biological anomalies (b) for the Mission II dataset. Example images of biological anomalies are shown in (a). Each image in the dataset was labeled with high, medium, or low perplexity (b). Although all three models do have differential responses in areas of high perplexity, the HDP model using standard features (e) outperforms the alternative models.

Authors: Genevieve Flaspohler, Nicholas Roy, Yogesh Girdhar