Identifying Plant Species with Herbarium Sheets.
The team once again joined PlantCLEF, an image-based plant identification competition. This year the objective involved a task of cross-domain classification between herbarium sheets and real-world photos. Herbarium sheets have been used by novices and experts alike to study and confirm plant species however, more research on its application in automated plant identification have yet to be carried out. Its visual appearance is very different from real-world images making it difficult in cross-domain classification. The team managed to place second and first in the first and second evaluation metric respectively.
This video explains the team's implementation and methods involved in PlantCLEF 2020. Our team presents the implementation and performance of a Herbarium-Field triplet loss network to evaluate the similarity of plant embeddings which corresponds to the cross-domain plant identification challenge in PlantCLEF 2020. A two-streamed triplet loss network was trained to maximise the embedding distance of different plant species and at the same time minimise the embedding distance of the same plant species. The team submitted seven runs which achieved a Mean Reciprocal Rank score of 0.121 and 0.111 for the whole test set and the sub-set of the test set respectively.
We will keep you update with our latest AI technology and applications