An Animal Detection Pipeline for Identification

Parham, Jason; Stewart, Charles; Crall, Jonathan; Rubenstein, Daniel I.; Holmberg, Jason; Berger-Wolf, Tanya Y.

doi:10.1109/wacv.2018.00123

Cited by 79 publications

(60 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…From a methodological perspective, a growing number of studies have utilized data mining and machine learning to classify illegal wildlife product sales (Hernandez-Castro & Roberts 2015, Harrison et al 2016, Austen et al 2018, Parham et al 2018. Our study used web scraping to simulate user searches on Facebook both retrospectively and prospectively.…”

Section: Type Of Post Number Of Posts Descriptionmentioning

confidence: 99%

The illegal wildlife digital market: an analysis of Chinese wildlife marketing and sale on Facebook

Cai

Mackey

2020

Envir. Conserv.

View full text Add to dashboard Cite

Summary At an estimated US$19 billion, the illicit wildlife trade is a serious threat to global conservation efforts. This criminal enterprise is now digital, expanding its footprint to consumers internationally by using the Internet and social media platforms. Recent studies have detected illegal wildlife selling posts on the popular social networking site Facebook in several different languages, including Chinese. In order to further explore this challenge to conservation, this study used big data approaches to identify and characterize wildlife trading activity in Chinese language on Facebook using an automated web scraper. We focused on keywords associated with elephants, rhinos and hawksbill turtles. We collected 10 303 unique Facebook posts over a 45-day period and were able to identify 639 posts from 268 unique users, which we suspect of directly marketing the sale of wildlife products. We also identified other species including Tibetan antelope, bears and African spurred tortoises. Facebook community pages appeared to have the highest percentage (48.2%) of wildlife selling posts. We also identified 14 different countries and regions with suspected wildlife-selling users, most located in Taiwan. Furthermore, we observed that the language used by some sellers changed from descriptive text to emojis and other code words. Collective action is needed from governments, law enforcement, civil society and technology companies leveraging big data approaches to better detect and interdict online Chinese-language wildlife trafficking.

show abstract

Section: Type Of Post Number Of Posts Descriptionmentioning

confidence: 99%

The illegal wildlife digital market: an analysis of Chinese wildlife marketing and sale on Facebook

Cai

Mackey

2020

Envir. Conserv.

View full text Add to dashboard Cite

show abstract

“…A range of cutting edges tools are now available to take advantage of CNN in the context of object detection for animal detection [32,33,34]. In particular, RetinaNet [26] is a CNN-based object detector able to detect a series of predefined object classes (e.g.…”

Section: Image Cropping With Cnn and Transfer Learningmentioning

confidence: 99%

Revisiting animal photo-identification using deep metric learning and network analysis

Miele

Dussert

Spataro

et al. 2020

Preprint

View full text Add to dashboard Cite

An increasing number of research programs rely on photographic capture-recapture (vs. direct marking) of individuals to study distribution and demography within animal populations. Photoidentification of individuals living in the wild is sometimes feasible using idiosyncratic coat or skin patterns, like for giraffes. When performed manually, the task is tedious and becomes almost impossible as populations grow in size. Computer vision techniques are an appealing and unavoidable help to tackle this apparently simple task in the big-data era. In this context, we propose to revisit giraffe re-identification using convolutional neural networks (CNNs).We first developed an end-to-end pipeline to retrieve a comprehensive set of re-identified giraffes from about 4, 000 raw photographs. To do so, we combined CNN-based object detection, SIFT pattern matching, and image similarity networks. We then quantified the performance of deep metric learning to retrieve the identity of known and unknown individuals. The re-identification performance of CNNs reached a top 5 accuracy of about 90%. Fully based on open-source software packages, our work paves the way for further attempts to build CNN-based pipelines for re-identification of individual animals, in giraffes but also in other species.

show abstract

“…Importantly, if new, automated matching techniques are developed and trained using a trusted baseline of data from thorough and systematic manual matching efforts, such as that presented here, these advanced techniques could replace hundreds or thousands of hours of visual curation with just a few automated hours of computation (Flynn et al, 2017). This is especially true if the fully automated pipelines can avoid the need for any pre-processing of imagery (Parham et al, 2018) before ensembled matching techniques use all available data to accurately match individual humpback whales across multiple, collaborating research sites along their range. Our work here can not only provide the high quality data and justification needed to train the next generation of multi-mark, automated matching systems (e.g.…”

Section: Implications For Development Of Algorithm Matching Technologmentioning

confidence: 99%

Photo-identification of individual Southern Hemisphere humpback whales (Megaptera novaeangliae) using all available natural marks:

Franklin

Brooks³

et al. 2020

Journal of Cetacean Research and Management

View full text Add to dashboard Cite

Misidentification errors in capture-mark recapture studies of humpback whales (Megaptera novaeangliae) related to poor quality of photographs as well as changes in natural marks can seriously affect population dynamics parameter estimates and derived estimates of population size when using sophisticated modelling techniques. In this study we used an innovative photo-identification matching system to investigate and examine the long-term stability and/or changes in natural marks on ventral-tail flukes, dorsal fin shapes and lateral body marks from a sample of 79 individual humpback whales, resighted in 2 to 11 years over timespans ranging from 2 to 21 years. A binary logistic mixed effects model, on a pair-matched sample of the 79 individual whales, found no significant differences in the proportions of ventral-tail fluke marks, dorsal fin shapes and lateral body marks, that displayed changes in primary and/or secondary characteristics over years (F=0.939, df=1/156, p =0.334). The results of this study substantiate the value and reliability of using primary and secondary natural marks on the ventral-tail flukes, in conjunction with dorsal fin shapes and secondary lateral body marks as double-tags. This provides a means of maximising observations of individual humpback whales over years, while minimising and managing misidentification errors in the photo-identification matching process, thus significantly improving modelling outcomes.

show abstract

An Animal Detection Pipeline for Identification

Cited by 79 publications

References 8 publications

The illegal wildlife digital market: an analysis of Chinese wildlife marketing and sale on Facebook

The illegal wildlife digital market: an analysis of Chinese wildlife marketing and sale on Facebook

Revisiting animal photo-identification using deep metric learning and network analysis

Photo-identification of individual Southern Hemisphere humpback whales (Megaptera novaeangliae) using all available natural marks:

Contact Info

Product

Resources

About