jrdurrant / specimen-vision

A set of tools for extracting metadata from biological specimen images
GNU General Public License v2.0
5 stars 3 forks source link

Computer Vision for biological specimen images

These are a set of tools for extracting metadata from specimen images. They include methods for determining specific properties of specimens, as well as general pre-processing tools.

Pre-processing

These can be used as a first step in making a specimen image more usable by automated methods.

Segmentation

Separates a specimen from the image background and produces a mask. This gives a silhouette that can be used for analysing the shape, as well as any other analysis that requires the background to be ignored.

Determining image scale

Given an image that contains a ruler, this method determines the scale that the image was taken at. This enables conversion between measurements in pixels and measurements in real world units.

Metadata

A set of methods for analysing specific properties. Currently this is limited to a set of butterflies and moths.

Morphometrics

Determining the area of the specimen's wings, as well as the length of the wing's leading edge.

Colour analysis

For an image, or a set of images, this finds the set of dominant colours. This can be helpful for comparison between species as well as simple visualisation.