Dataset for Hop varieties classification

Abstract

Humulus lupulus L., also known as hops, is a vine whose flowers are a major component in brewing. It delivers flavor, bitterness, and aroma to beer and also aids in foam stabilization. Furthermore, it plays an important role in beer conservation due to its antimicrobial and antioxidant properties, which have recently been studied for food preservation. Hops can also be found in the production of cosmetics and is considered healthy food. There are more than 250 cataloged varieties of hops, and among the main attributes that differ from each other are alpha-acids, beta-acids, and essential oils. Those components give the beer a unique combination of characteristics, and may even influence its category. There are many ways to identify the hop variety from its acids and essential oils using methods such as chromatography, mass spectrometry, capillary electrophoresis, and nuclear magnetic resonance. However, these methods demand expensive and complex equipment, inaccessible or unavailable to most beer producers. In this work, we present a database that includes 1592 images of hop leaves, from 12 popular hop varieties in southeastern Brazil. From these images, it is possible to explore methods of pattern recognition and machine learning to classify hop varieties.

Keywords: Hop varieties; Leaf recognition; Plant recognition.

Fig. 1

Examples of the 12 Hops varieties contained in this data set: (a) Cascade; (b) Nugget; (c) Cluster; (d) Triple Pearl; (e) Hallertau Mittelfrueh; (f) Centennia; (g) Saaz; (h) Sorachi Ace; (i) Tahoma; (j) Comet; (k) Triumph; (l) Zeus.

Fig. 2

Labeling methodology: leaf with the largest area was labeled as main (red bounding box), while the others as extra leaves (yellow).