LifeCLEF 2015: Multimedia Life Species Identification Challenges

Abstract

Abstract. Using multimedia identification tools is considered as one of the most promising solutions to help bridge the taxonomic gap and build accurate knowledge of the identity, the geographic distribution and the evolution of living species. Large and structured communities of nature observers (e.g., iSpot, Xeno-canto, Tela Botanica, etc.) as well as big monitoring equipment have actually started to produce outstanding collections of multimedia records. Unfortunately, the performance of the state-of-the-art analysis techniques on such data is still not well understood and is far from reaching real world requirements. The LifeCLEF lab proposes to evaluate these challenges around 3 tasks related to multimedia information retrieval and fine-grained classification problems in 3 domains. Each task is based on large volumes of real-world data and the measured challenges are defined in collaboration with biologists and environmental stakeholders to reflect realistic usage scenarios. For each task, we report the methodology, the data sets as well as the results and the main outcomes.

How deep learning extracts and learns leaf features for plant classification

“…This is still far from matching the scale and variety of existing general major datasets for images [52,55,51], videos [53] or languages [54]. In addition, we can see that the PlantClef2015 dataset [6] has one of the largest number of object categories but the least number of images. For example, compared to the ILSVRC 2010 dataset [55], it has less than 10 % of their total images but the M A N U S C R I P T same number of categories.…”

“…Classification performance of the network trained with all training sets (w = 2,288, P = 34,672) is obviously better compared to that trained on smaller subset of data (W = 1,324, P = 3,960). Next, although reducing CNN layer depth might affect A C C E P T E D M A N U S C R I P T [52] 8.3 milion 365 Sport-1M [53] 1 milion 487 Visual Genome QA [54] 1.7 million questions/answer pairs -ILSVRC 2010 [55] 1.4 milion 1000 PlantClef2015 dataset [6] 113,205 1000…”

“…Table 6 shows examples of existing well-known datasets and their size as quantity of images. The biggest plant database that we have found is the PlantClef2015 dataset [6] which has only around 113,205 number of images. This is still far from matching the scale and variety of existing general major datasets for images [52,55,51], videos [53] or languages [54].…”

“…A number of approaches have been proposed in the literature for automatic analysis of botanical organs, such as leaves and flowers [4,5,6]. In botany, leaves are almost always used to supply important diagnostic characters for plant classification and in some groups exclusively so.…”

Coral Reef Fish Detection and Recognition in Underwater Videos by Supervised Machine Learning: Comparison Between Deep Learning and HOG+SVM Methods

“…In the 2015 SeaClef contest [23], the best results have shown that Deep Learning can achieve a better classification for fish detection than SVM or other classical methods. This may be due to the fact that in Deep Learning, features are automatically built by the classifier itself, in an optimal way.…”