Building accurate knowledge of the identity, the geographic distribution and the evolution of species is essential for the sustainable development of humanity, as well as for biodiversity conservation. However, the difficulty of identifying plants and animals is hindering the aggregation of new data and knowledge. Identifying and naming living plants or animals is almost impossible for the general public and is often difficult even for professionals and naturalists. Bridging this gap is a key step towards enabling effective biodiversity monitoring systems. The LifeCLEF campaign, presented in this paper, has been promoting and evaluating advances in this domain since 2011. The 2021 edition proposes four data-oriented challenges related to the identification and prediction of biodiversity: (i) PlantCLEF: cross-domain plant identification based on herbarium sheets, (ii) BirdCLEF: bird species recognition in audio soundscapes, (iii) GeoLifeCLEF: remote sensing based prediction of species, and (iv) SnakeCLEF: Automatic Snake Species Identification with Country-Level Focus.

/pdf/overview-of-lifeclef-2021-an-evaluation-of-machine-learning-29nkwexydi.pdf

Overview of LifeCLEF 2021: An Evaluation of Machine-Learning Based Species Identification and Species Distribution Prediction

It is estimated that there are more than 300,000 species of vascular plants in the world. Increasing our knowledge of these species is of paramount importance for the development of human civilization (agri-culture, construction, pharmacopoeia, etc.), especially in the context of the biodiversity crisis. However, the burden of systematic plant identification by human experts strongly penalizes the aggregation of new data and knowledge. Since then, automatic identification has made considerable progress in recent years as highlighted during all previous editions of PlantCLEF. Deep learning techniques now seem mature enough to address the ultimate but realistic problem of global identification of plant biodiversity in spite of many problems that the data may present (a huge number of classes, very strongly unbalanced classes, partially erroneous identifications, duplications, variable visual quality, diversity of visual contents such as photos or herbarium sheets, etc). The PlantCLEF2022 challenge edition proposes to take a step in this direction by tackling a multi-image (and metadata) classification problem with a very large number of classes (80k plant species). This paper presents the resources and evaluations of the challenge, summarizes the approaches and systems employed by the participating research groups, and provides an analysis of key findings.

Overview of PlantCLEF 2022: Image-based plant identification at global scale

ImageCLEF is a part of the Conference and Labs of the Evaluation Forum (CLEF) initiative and includes a variety of tasks dedicated to multimodal image information retrieval, including image classification and annotation. The tuberculosis (TB) task is one of the ImageCLEF tasks that started in 2017 and changed from year to year with more complex challenges. The 2022 edition was dedicated to the automatic analysis of caverns, and included two subtasks: cavern region detection and cavern reporting. In 2022, 6 groups from 5 countries submitted at least one successful run for at least one subtask. This paper describes the TB task setup, data and the approaches of the participants. The team approach was based on a public pretrained 3D lung nodule detection model.

Overview of ImageCLEFtuberculosis 2022 - CT-based Cavern Detection and Report

The 2022 ImageCLEFfusion task is the first edition of this task, targeting the creation of late fusion or ensembling methods in two different scenarios: (i) the prediction of media visual interestingness, and (ii) social media image search results diversification. The objective proposed to participants is to train and test their proposed fusion schemes on a set of pre-computed inducers, without creating or bringing inducers from the outside. The two scenarios correspond to a regression scenario in the case of media interestingness, where performance is measured via the mean average precision at 10 (MAP@10) metric, and to a retrieval scenario in the case of result diversification, where performance is measured via the F1-score and Cluster Recall at 20 (F1@20, CR@20). Overall 6 teams registered for ImageCLEFfusion, 5 of them submitting runs, while only one team submitted runs to both the interestingness and diversification tasks. A total of 39 runs were received, and an analysis of the proposed methods shows a great diversity among them, ranging from statistical weighted approaches, weighted approaches that use learning stages for creating the weights, machine learning approaches that join the inducer predictions like SVM or KNN, deep learning approaches, and even fusion schemes that join the results of other fusion schemes.

Overview of ImageCLEFfusion 2022 Task - Ensembling Methods for Media Interestingness Prediction and Result Diversification

Cloud computing is an inevitable technology that leads to demanding business requirements, the cloud services of IaaS utilization are being expanded in the data storage space, which leads to an increase in the requirements of bandwidth. To reduce the storage space, the De-Duplication Compression (DDC) algorithm has been proposed to improve the storage and utilization of bandwidth in the Storage optimization System (SoS) through the elimination of duplicate files and monitor the IaaS storage utilization. The various types of metadata standards help to identify the respective file objects and the file data elements are linked with corresponding block can be clustering into segmented bins. The DDC algorithm is associated with a linked list structure to access the file on different parameters to classify the future file access patterns against the cloud storage. Based on user access frequency, the ranking has been determined to predict future usage based on file access patterns, the SoS system possesses a dashboard that can help the user to decide on the file operation. This system result has experimented with the real environment with different parameter simulation. The proposed SoS framework will decrease cloud IaaS space size up to 13.80% than the existing cloud storage system.

Enhanced Storage optimization System (SoS) for IaaS Cloud Storage

The healthcare industry, through digitization, is trying to achieve interoperability, but has not been able to achieve complete Health Information Exchange (HIE). One of the major challenges in achieving this is the inability to accurately match patient data. Mismatching of patient records can lead to improper treatment which can prove to be fatal. Also, the presence of duplicate overheads has caused inaccessibility to crucial information in the time of need. Existing solutions to patient matching are both time-consuming and non-scalable. This paper proposes a framework, namely, Electronic Patient Matching System (EPMS), which attempts to overcome these barriers while achieving a good accuracy in matching patient records. The framework encodes the patient records using variational autoencoder and amalgamates them by performing locality sensitive hashing on an Apache spark cluster. This makes the process faster and highly scalable. Furthermore, a fuzzy matching of the records in each block is performed using Levenshtein distances to identify the duplicate patient records. Experimental investigations were performed on a synthetically generated dataset consisting of 44555 patient records. The proposed framework achieved a matching accuracy of 81.15% on this dataset.

EPMS: A Framework for Large-Scale Patient Matching

Plant classification is the process of identifying plant species using image processing techniques. This system is helpful in identifying the different plant species which has useful applications in botanical studies and many plant based industries. The previous works have built systems based on data sets containing limited number of classes . However, the plant classification systems can further be improved in order to cover more number of species. The proposed system is capable of identifying 80000 classes of plant species and is built with a large dataset using AlexNet deep learning architecture. A combination of AdaGrad and KL Divergence optimization and loss functions respectively is used to train the model that produced a MRR score of 0.00029.

Classification of Plant Species Using AlexNet Architecture

In a cloud storage, while deduplication enables optimal storage space utilization, it also incurs a substantial overhead in maintaining the metadata, namely, the Fingerprint Index and the File Recipe. As this metadata is huge, it has to be stored in the disk which causes considerable read-latency in Deduplicated Cloud Storage (DCS). In order to improve this read-latency, it would be highly beneficial to prefetch the relevant fingerprints in a cache. Many existing research solutions have utilized either the spatial locality or the similarity among the files to prefetch the relevant fingerprints. However, the DCS that has been designed and implemented in this paper is intended to cater to non-backup workloads that do not exhibit significant spatial locality or similarity among the files. Hence, this paper proposes a suitable alternative prefetching approach that mines the pattern of client read accesses to find the most frequently accessed files. The proposed prefetching approach has been implemented and incorporated in the DCS. The experimental investigations indicate that the proposed prefetching approach improves the cache hit rates by 140% and increases the read throughput by 88% when compared with the Extreme Binning approach (Bhagwat, D., Eshghi, K. Long, D.D.E. and Lillibridge, M. (2009) Extreme Binning: Scalable, Parallel Deduplication for Chunk-based File Backup. Proc. MASCOTS'09, London, UK, September 21–23, pp. 1–9. IEEE) while incurring only a marginal computational overhead of 1.7 s.

Improving Read Throughput of Deduplicated Cloud Storage using Frequent Pattern-Based Prefetching Technique

Result diversification provides a broader view of a search topic while also increasing the possibilities of finding relevant information. It has shown to increase user satisfaction in recommender systems and web search. Many different approaches have been proposed in the related literature for the diversification problem. Since the web search result is huge in size, it is essential to have an efficient fusion approach. Hence, objective of this paper is to propose the implementation of fusion model based on KNN, CART and SVR regressors. This fusion model aims to improve the accuracy and reduce the error value of the result that is generated.

A Fusion Approach for Web Search Result Diversification using Machine Learning Algorithms

Result diversification provides a broader view of a topic, while maximizing the chances of retrieving relevant information. It avoids the bias in results, thus improving the user experience. This area finds a lot of applications in web searches and recommendation systems. The existing literature on this domain has achieved good accuracy on smaller datasets and using single models. An ensemble approach, using three neural network models, has been proposed to improve the existing predictions using a bigger dataset.

Prabavathy Balasundaram

Papers

EPMS: A Framework for Large-Scale Patient Matching

Classification of Plant Species Using AlexNet Architecture

Improving Read Throughput of Deduplicated Cloud Storage using Frequent Pattern-Based Prefetching Technique

A Fusion Approach for Web Search Result Diversification using Machine Learning Algorithms

Ensembled Approach for Web Search Result Diversification Using Neural Networks