Categorizing plant images at the variety level: Did you say fine-grained?

Abstract

International audienceThis paper addresses the problem of categorizing plant images at the variety level, i.e. at a finer taxonomic grain than state-of-the-art studies usually working at the species level. It therefore introduces two new evaluation datasets of agro-biodiversity interest, each being related to concrete scenarios on large-scale plant resources. They have been chosen so as to involve very different acquisition protocols and visual patterns in order to evaluate if state-of-the-art image classification techniques can generalize to such specific contexts and avoid the cost of building specific ad-hoc solutions. The first one is a collection of 2071 pictures of loose rice seeds built from 95 accessions kept in a bank of seeds. The second one is a collection of 2037 pictures of grape leaves taken in the fields and belonging to 34 varieties among the most commonly ones used in viticulture. Both datasets exhibit a very low inter-class variability resulting in two challenging fine-grained classification tasks, even for expert human operators. A baseline experimental study was conducted on the two datasets using the two most effective families of classification techniques in the state-of-the-art, i.e. convolutional neural networks on one side and fisher vectors-based discriminant models on the other side. It shows that the achieved classification performance is very different between the two problems. It is actually pretty bad for the grape leaves collection but much better in the case of the rice seeds collection for which the acquisition protocol was much more constrained and the morphological variability more visible. The conclusion is that automatically identifying plant varieties might already be feasible for some specific scenarios and in controlled environments but that it is still an open problem in the general case

Trends and gaps in the use of citizen science derived data as input for species distribution models: A quantitative review

“…Considering the known number of species in each group according to the Catalogue of Life database (vascular plants: 348,000 species; fungus: 140,000 species; bryophytes 16,000 species), plant taxonomic groups were remarkably under-represented in CS papers (n = 13; 8% of all groups in this review). The major obstacle could be that identifying plants up to species level in the field is sometimes complex, even for expert botanists [ 73 , 74 ]. Plant identification is time consuming for several families, requires significant botanical skills, and can be frustrating for non-expert volunteers [ 74 ].…”

Identification of wheat kernels by fusion of RGB, SWIR, and VNIR samples

“…Eventually, it reveals the four-orthonormal basis, (z 1 , z 2 , z 3 , z 4 ) spanning the difference subspace of the processed channels. • To obtain the common vector, a com , the reference vector is projected onto the orthonormal basis and a difference vector a diff is formed, which is represented in Eqn (7). Once difference vector is subtracted from the reference vector, the common vector, a com , would be modeled as shown in Eqn (8).…”

“…The 88.33% accuracy was determined using traditional dense-scale invariant features (DSIFT) and an SVM classifier. In a study by Champ et al,7 a classification score of 88.67% was achieved using a CNN and two plant species datasets: (i) 2071 loose rice seed of 95 varieties and (ii) a collection of 2037 grape leaves belonging to the 34 most common varieties used in viticulture. Two feature-extraction approaches (CNNs and vector discriminant models (VDMs)) were used to identify grapes by their leaves.…”

Advances in site‐specific weed management in agriculture—A review

“…Site‐specific weed management started with grid sampling and weed mapping in the late 20th century (Marshall, 1988; Wiles et al, 1992). Current advances of this concept comprise several commercial and practical applications in camera‐guided weed hoeing, sensor‐based patch and spot spraying, robotic weeding and weed scouting (Champ et al, 2016; Dyrmann et al, 2016; Gutjahr et al, 2012; Kunz et al, 2015; Peteinatos et al, 2020; Tillett et al, 2002). Several more prototypes, such as sensor‐based electrical weed control (Reiser et al, 2019), Unmanned Aerial Vehicle (UAV)‐based weed mapping and patch spraying (Mink et al, 2018; Rasmussen et al, 2013) and targeted weed control with direct injection of herbicides (Pohl et al, 2019; Ruigrok et al, 2020), are currently being developed.…”