• +44 (0115) 84 66590
  • mercedestorrestorres@nottingham.ac.uk

DATABASES

Here you can find information about the Habitat K databases: Habitat 1K and Habitat 3K.

Both databases were collected as part of my PhD research, which revolved around Automatic Phase 1 Habitat Classification using ground-taken imagery.

For my research, I defined a ground-taken photograph as a digital photograph taken by a human on the ground. The subject of the photograph must be at least one habitat. There are no limitations to the number of habitat classes within a photograph.

Additionally, there are no limitations to its layout in terms of orientation or perspective. These photographs may be taken with any type of digital or mobile camera and may or may not be geo-referenced. In the case of the Habitat K databases, all images are georeferenced.

Both databases were classified by a Phase-1 trained expert and annotated by me.

Annotations

The annotations, which you can download from the Databases section, are stored in XML documents. They contain information about the habitats present in each image and their localisation within each image. All annotations follow the guidelines from the Phase 1 classification scheme, with up to three levels included. Figure 1 shows a two screen captures of two different annotated images.

Figure 1. Anotated Images using Korc and Schneider's Annotation Tool for MATLAB.

I used the Annotation Tool for MATLAB developed by Filip Korc and David Schneider. If you want to visualise or edit the annotations, you can download it from here.

Medium-Level Annotations

As part of my PhD research, I studied the effect of including annotations in the decision process made from non-experts. These annotations are much more simple in nature (i.e, instead of having a class for 'coniferous leaves', 'broad-leaved trees' and 'mixed-leaves trees' , we have only two classes: 'trees with tress' and 'trees without trees') and provided a significant increase in accuracy in our system.

The full list of medium-level annotations is: Trees - leaves, Trees - no leaves, Trees - mixed leaves, Scrub, Grass - with flowers, Grass - uniform, Grass - reed, Bracken or fern, Tall herb, Heath, Water - runing, Arable land, Boundary - scrub/trees, Wall, Fence, Sky, Grass - bright green, Grass - not green, Sand - mud, Small rocks, Big rocks, Water - standing , Cliff - water, Cliff - no water , Winter, Autum, Summer, Spring, Brown, Yellow, Red, White, Blue, Green