Finding Nemo’s Giant Cousin: Keypoint Matching for Robust Re-Identification of Giant Sunfish

Pedersen, Malte; Nyegaard, Marianne; Moeslund, Thomas B.

doi:10.3390/jmse11050889

Cited by 2 publications

(1 citation statement)

References 59 publications

(86 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Obtaining sufficiently large datasets for elusive species such as molids can be challenging. An alternative approach that shows promising preliminary results for M. alexandrini combines traditional use of algorithms and machine learning, without the need for prior training , Pedersen et al, 2023. Further development of this method could increase match capacity, and be applied to other molid hotspots such as the Galapagos Islands (www.oceansunfish.org), and perhaps aid in photo identification of other elusive taxa.…”

Section: Automated Recognitionmentioning

confidence: 99%

Rapid physiological colouration change is a challenge - but not a hindrance - to successful photo identification of giant sunfish (Mola alexandrini, Molidae)

Nyegaard,

Karmy,

McBride

et al. 2023

Front. Mar. Sci.

Self Cite

View full text Add to dashboard Cite

Photo ID is a common tool in ecology, but has not previously been attempted for the ocean sunfishes (Mola spp., Molidae; ‘molids’). The technique, based on body patterns, could potentially be informative for studying the seasonal occurrence of giant sunfish (Mola alexandrini) on the Bali reefs (Indonesia), where this species is an important drawcard for the local SCUBA diving tourism. However, molids are capable of rapid physiological colouration change, which may complicate the application of the method. Our study aimed to determine if photo ID is nevertheless achievable and informative. To test this, we created the citizen-science platform ‘Match My Mola’ for crowd-sourcing imagery (photos and video) of M. alexandrini in Bali, and undertook trial matching (n=1,098 submissions). The submitted imagery revealed a wide range of pattern clarity, from fish with no pattern to bold displays. Video confirmed physiological colouration change can occur in seconds in this species from low to high contrast, and cause individuals to look very different between moments. However, individual patterns appear to be stable although at least some parts can become inconspicuous during low contrast displays. Despite of this, photo ID is possible, including in some instances, where only partial patterns are visible on one image compared with another. However, true negatives (confirming two fish are not the same) can be challenging. Most identified matches were of fish photographed by different divers on the same day. Only a small number (n=9) were found with resighting durations ≥1 day (1 – 2,652 days). These matches demonstrate that at least some individuals return to the same reefs both within and between seasons, with the resighting duration of 7.2 years constituting the longest known example of molid site fidelity. Comparing body morphology between resightings of > 1 year (n=6) revealed limited indications of growth, contradicting the current understanding of rapid growth in captive molids (Mola mola), and highlighting the knowledge gap regarding growth in the wild. Continued photo ID in the Bali area could provide valuable complementary information to future growth studies using other methods as well as provide further insights into molid site fidelity.

show abstract

Section: Automated Recognitionmentioning

confidence: 99%

Rapid physiological colouration change is a challenge - but not a hindrance - to successful photo identification of giant sunfish (Mola alexandrini, Molidae)

Nyegaard,

Karmy,

McBride

et al. 2023

Front. Mar. Sci.

Self Cite

View full text Add to dashboard Cite

show abstract

Species-Agnostic Patterned Animal Re-identification by Aggregating Deep Local Features

Nepovinnykh,

Chelak,

Eerola

et al. 2024

Int J Comput Vis

View full text Add to dashboard Cite

Access to large image volumes through camera traps and crowdsourcing provides novel possibilities for animal monitoring and conservation. It calls for automatic methods for analysis, in particular, when re-identifying individual animals from the images. Most existing re-identification methods rely on either hand-crafted local features or end-to-end learning of fur pattern similarity. The former does not need labeled training data, while the latter, although very data-hungry typically outperforms the former when enough training data is available. We propose a novel re-identification pipeline that combines the strengths of both approaches by utilizing modern learnable local features and feature aggregation. This creates representative pattern feature embeddings that provide high re-identification accuracy while allowing us to apply the method to small datasets by using pre-trained feature descriptors. We report a comprehensive comparison of different modern local features and demonstrate the advantages of the proposed pipeline on two very different species.

show abstract

Finding Nemo’s Giant Cousin: Keypoint Matching for Robust Re-Identification of Giant Sunfish

Cited by 2 publications

References 59 publications

Rapid physiological colouration change is a challenge - but not a hindrance - to successful photo identification of giant sunfish (Mola alexandrini, Molidae)

Rapid physiological colouration change is a challenge - but not a hindrance - to successful photo identification of giant sunfish (Mola alexandrini, Molidae)

Species-Agnostic Patterned Animal Re-identification by Aggregating Deep Local Features

Contact Info

Product

Resources

About