scholarly journals FIN-PRINT a fully-automated multi-stage deep-learning-based framework for the individual recognition of killer whales

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Christian Bergler ◽  
Alexander Gebhard ◽  
Jared R. Towers ◽  
Leonid Butyrev ◽  
Gary J. Sutton ◽  
...  
Keyword(s):  
Deep Learning ◽  
North America ◽  
Killer Whale ◽  
Individual Recognition ◽  
Killer Whales ◽  
Photo Identification ◽  
Learning Framework ◽  
Multi Stage ◽  
The Individual ◽  
Identification Techniques

AbstractBiometric identification techniques such as photo-identification require an array of unique natural markings to identify individuals. From 1975 to present, Bigg’s killer whales have been photo-identified along the west coast of North America, resulting in one of the largest and longest-running cetacean photo-identification datasets. However, data maintenance and analysis are extremely time and resource consuming. This study transfers the procedure of killer whale image identification into a fully automated, multi-stage, deep learning framework, entitled FIN-PRINT. It is composed of multiple sequentially ordered sub-components. FIN-PRINT is trained and evaluated on a dataset collected over an 8-year period (2011–2018) in the coastal waters off western North America, including 121,000 human-annotated identification images of Bigg’s killer whales. At first, object detection is performed to identify unique killer whale markings, resulting in 94.4% recall, 94.1% precision, and 93.4% mean-average-precision (mAP). Second, all previously identified natural killer whale markings are extracted. The third step introduces a data enhancement mechanism by filtering between valid and invalid markings from previous processing levels, achieving 92.8% recall, 97.5%, precision, and 95.2% accuracy. The fourth and final step involves multi-class individual recognition. When evaluated on the network test set, it achieved an accuracy of 92.5% with 97.2% top-3 unweighted accuracy (TUA) for the 100 most commonly photo-identified killer whales. Additionally, the method achieved an accuracy of 84.5% and a TUA of 92.9% when applied to the entire 2018 image collection of the 100 most common killer whales. The source code of FIN-PRINT can be adapted to other species and will be publicly available.

ORCA-SLANG: An Automatic Multi-Stage Semi-Supervised Deep Learning Framework for Large-Scale Killer Whale Call Type Identification

2021 ◽  
Author(s):  
Christian Bergler ◽  
Manuel Schmitt ◽  
Andreas Maier ◽  
Helena Symonds ◽  
Paul Spong ◽  
...  
Keyword(s):  
Deep Learning ◽  
Large Scale ◽  
Killer Whale ◽  
Call Type ◽  
Learning Framework ◽  
Multi Stage

An Enhanced Multi-Stage Deep Learning Framework for Detecting Malicious Activities From Autonomous Vehicles

2021 ◽  
pp. 1-10
Author(s):  
Izhar Ahmed Khan ◽  
Nour Moustafa ◽  
Dechang Pi ◽  
Waqas Haider ◽  
Bentian Li ◽  
...  
Keyword(s):  
Deep Learning ◽  
Autonomous Vehicles ◽  
Learning Framework ◽  
Multi Stage

Prey switching by killer whales in the north-east Atlantic: observational evidence and experimental insights

2013 ◽  
Vol 94 (6) ◽  
pp. 1357-1365 ◽  
Author(s):  
Dag Vongraven ◽  
Anna Bisther
Keyword(s):  
Killer Whale ◽  
Large Population ◽  
Tooth Wear ◽  
Killer Whales ◽  
Prey Switching ◽  
Photo Identification ◽  
North East ◽  
The North ◽  
Spawning Stock

Studies in the Pacific have identified distinct killer whale ecotypes that are either specialized mammal- or fish-eaters. The different types have developed hunting strategies that would suggest specialization could be more advantageous than generalism. However, it has been suggested, based on long-term dietary markers of tooth wear and stable isotope values, that lineages in the North Atlantic are generalist, but with individual variation in the proportion of prey types consumed. Here, we present the results of ten years of observational and photo-identification data of a population of killer whales that follows the Norwegian spring-spawning stock of Atlantic herring. Although the whales were predominantly observed while feeding upon herring, one pod of herring-eating whales was also observed interacting with seals. This supports the hypothesis based on the long-term markers, of a degree of specialization, with a small number of groups persistently feeding upon mammals, but switching between herring and seals. We further investigated this prey switching by conducting playbacks of herring-eating killer whale sounds to harbour seals at haul-out sites on the herring spawning grounds. We recorded changes in behaviour consistent with an anti-predator response, suggesting the seals perceived the herring-eating killer whales as a potential predatory threat and had not habituated to their calls. This could be due to the risk of herring-eating killer whales switching to mammalian prey, or the difficulty of discriminating between killer whale pods due to the large population size and number of killer whale call dialects in this population, or a combination of both.

Using opportunistic photo-identifications to detect a population decline of killer whales (Orcinus orca) in British and Irish waters

2013 ◽  
Vol 94 (6) ◽  
pp. 1327-1333 ◽  
Author(s):  
Suzanne Beck ◽  
Andrew D. Foote ◽  
Sandra Kötter ◽  
Olivia Harries ◽  
Laura Mandleberg ◽  
...  
Keyword(s):  
Killer Whale ◽  
Population Decline ◽  
Small Population ◽  
High Ratio ◽  
Older Individuals ◽  
Adult Males ◽  
Killer Whales ◽  
Demographic Stochasticity ◽  
Photo Identification ◽  
Bayesian Formulation

An assemblage of killer whales that has been sighted in waters off the west coast of the British Isles and Ireland has previously been shown to be isolated from other North Atlantic killer whale communities based on association patterns. By applying a Bayesian formulation of the Jolly–Seber mark-recapture model to the photo-identification data compiled from opportunistic photographic encounters with this population of killer whales, we show that such sparse and opportunistically-collected data can still be valuable in estimating population dynamics of small, wide-ranging groups. Good quality photo-identification data was collected from 32 encounters over 19 years. Despite a cumulative total of 77 identifications from these encounters, just ten individuals were identified and the remaining 67 identifications were re-sights of these ten animals. There was no detected recruitment through births during the study and, as a result, the population appears to be in a slight decline. The demography of the population was highly skewed towards older individuals and had an unusually high ratio of adult males, and we suggest that demographic stochasticity due to a small population size may be further impacting the population growth rate. We recommend that this population be managed as a separate conservation unit from neighbouring killer whale populations.

scholarly journals Efficient Stain-Aware Nuclei Segmentation Deep Learning Framework for Multi-Center Histopathological Images

Electronics ◽  
2021 ◽  
Vol 10 (8) ◽  
pp. 954
Author(s):  
Loay Hassan ◽  
Mohamed Abdel-Nasser ◽  
Adel Saleh ◽  
Osama A. Omer ◽  
Domenec Puig
Keyword(s):  
Deep Learning ◽  
Choquet Integral ◽  
Adjacent Cell ◽  
Cell Nuclei ◽  
Nuclei Segmentation ◽  
Learning Framework ◽  
Segmentation Methods ◽  
Histopathological Images ◽  
The Individual ◽  
Whole Slide Images

Existing nuclei segmentation methods have obtained limited results with multi-center and multi-organ whole-slide images (WSIs) due to the use of different stains, scanners, overlapping, clumped nuclei, and the ambiguous boundary between adjacent cell nuclei. In an attempt to address these problems, we propose an efficient stain-aware nuclei segmentation method based on deep learning for multi-center WSIs. Unlike all related works that exploit a single-stain template from the dataset to normalize WSIs, we propose an efficient algorithm to select a set of stain templates based on stain clustering. Individual deep learning models are trained based on each stain template, and then, an aggregation function based on the Choquet integral is employed to combine the segmentation masks of the individual models. With a challenging multi-center multi-organ WSIs dataset, the experimental results demonstrate that the proposed method outperforms the state-of-art nuclei segmentation methods with aggregated Jaccard index (AJI) and F1-scores of 73.23% and 89.32%, respectively, while achieving a lower number of parameters.

Population size, survival and reproductive rates of northern Norwegian killer whales (Orcinus orca) in 1986–2003

2013 ◽  
Vol 94 (6) ◽  
pp. 1277-1291 ◽  
Author(s):  
Sanna Kuningas ◽  
Tiu Similä ◽  
Philip S. Hammond
Keyword(s):  
Population Size ◽  
Killer Whale ◽  
Survival Rates ◽  
Clupea Harengus ◽  
Orcinus Orca ◽  
Adult Males ◽  
Killer Whales ◽  
Apparent Survival ◽  
Photo Identification ◽  
Reproductive Rates

A long-term photo-identification study of killer whales (Orcinus orca) in northern Norway was initiated in 1986, when their prey the Norwegian spring-spawning herring (Clupea harengus) started to winter in a complex fjord system. The aim of this work was to estimate population size and apparent survival rates in this killer whale population using photo-identification and mark–recapture techniques with data collected during October–December 1986–2003. Total population size was estimated to be highest in 2003: 731 individuals (SE = 139, 95% CI = 505–1059) using a model taking heterogeneity of capture probabilities into account. Apparent survival of adult males and adult females was estimated using the Cormack–Jolly–Seber model as 0.971 (SE = 0.008) and 0.977 (SE = 0.009), respectively. Calving intervals ranged from 3 to 14 years (mean = 5.06, SE = 0.722). These are the first estimates of northern Norwegian killer whale population parameters, allowing their dynamics to be investigated and comparisons to be made with killer whale populations globally.

scholarly journals Aggregation Tool for Genomic Concepts (ATGC): A deep learning framework for sparse genomic measures and its application to tumor mutational burden

2020 ◽  
Author(s):  
Jordan Anaya ◽  
John-William Sidhom ◽  
Craig A. Cummings ◽  
Alexander S. Baras ◽  
Keyword(s):  
Deep Learning ◽  
Large Scale ◽  
Multiple Instance Learning ◽  
Machine Learning Algorithms ◽  
Learning Framework ◽  
Mutational Burden ◽  
Tumor Mutational Burden ◽  
Training Examples ◽  
The Individual ◽  
Gene Panels

ABSTRACTDeep learning has the ability to extract meaningful features from data given enough training examples. Large scale genomic data are well suited for this class of machine learning algorithms; however, for many of these data the labels are at the level of the sample instead of at the level of the individual genomic measures. To leverage the power of deep learning for these types of data we turn to a multiple instance learning framework, and present an easily extensible tool built with TensorFlow and Keras. We show how this tool can be applied to somatic variants (featurizing genomic position and sequence context), and accurately classify samples according to whether they contain a specific variant (hotspot or tumor suppressor) or whether they contain a type of variant (microsatellite instability). We then apply our model to the calibration of tumor mutational burden (TMB), an increasingly important metric in the field of immunotherapy, across a variety of commonly used gene panels. Regardless of the panel, we observed improvements in regression to the gold standard whole exome derived value for this metric, with additional performance benefits as more data were provided to the model (such as noncoding variants from panel assays). Our results suggest this framework could lead to improvements in a range of tasks where the sample level metric is determined by the aggregation of a set of genomic measures, such as somatic mutations that we focused on in this study.

Iris photo-identification: A new methodology for the individual recognition of Tarentola geckos

2013 ◽  
Vol 34 (4) ◽  
pp. 590-596 ◽  
Author(s):  
Ricardo Rocha ◽  
Tiago Carrilho ◽  
Rui Rebelo
Keyword(s):  
Pattern Matching ◽  
Individual Recognition ◽  
Field Studies ◽  
Individual Identification ◽  
Identification System ◽  
Correct Identification ◽  
Photo Identification ◽  
Horizontal Angle ◽  
The Individual

Gekkonid field studies are hampered by the difficulty to individually recognize individuals. In this study we assess the feasibility of using their variegated iris pattern to photo-identify Tarentola boettgeri bischoffi, a threatened Macaronesian endemic. Using a library of 924 photos taken over a 9-month period we also evaluate the use of the pattern matching software Interactive Individual Identification System (I3S) to match photos of known specimens. Individuals were clearly recognized by their iris pattern with no misidentifications, and using I3S lead to a correct identification of 95% of the recaptures in a shorter time than the same process when conducted visually by an observer. The method’s feasibility was improved by increasing the number of images of each animal in the library and hindered by photos that deviate from a horizontal angle.

A Deep Learning-Based Framework for Accurate Facial Ethnicity Classification and Efficient Query Retrieval

2021 ◽  
pp. 216-240
Author(s):  
Geraldine Amali ◽  
Keerthana K. S. V. ◽  
Jaiesh Sunil Pahlajani
Keyword(s):  
Deep Learning ◽  
Search Time ◽  
Suitable Model ◽  
Retrieval Model ◽  
Learning Framework ◽  
Learning Techniques ◽  
And Gender ◽  
The Individual ◽  
Ethnicity And Gender ◽  
Ethnicity Classification

Facial images carry important demographic information such as ethnicity and gender. Ethnicity is an essential part of human identity and serves as a useful identifier for numerous applications ranging from biometric recognition, targeted advertising to social media profiling. Recent years have seen a huge spike in the use of convolutional neural networks (CNNs) for various visual, face recognition problems. The ability of the CNN to take advantage of the hierarchical pattern in data makes it a suitable model for facial ethnicity classification. As facial datasets lack ethnicity information it becomes extremely difficult to classify images. In this chapter a deep learning framework is proposed that classifies the individual into their respective ethnicities which are Asian, African, Latino, and White. The performances of various deep learning techniques are documented and compared for accuracy of classification. Also, a simple efficient face retrieval model is built which retrieves similar faces. The aim of this model is to reduce the search time by 1/3 of the original retrieval model.

Sign in / Sign up

Export Citation Format

Share Document