A method for resolving occlusions when multitracking individuals in a shoal

Dolado, Ruth; Gimeno, Elisabet; Beltran, Francesc S.; Quera, Vicenç; Pertusa, José

doi:10.3758/s13428-014-0520-9

Cited by 19 publications

(22 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We proceed to the next iteration by considering F ρ (2) . If one of the aforementioned conditions is not satis ed, the individual fragment is marked as non-consistent as well as the entire global fragment G.…”

Section: D43 Protocol 1: One-global-fragment Trackingmentioning

confidence: 99%

See 1 more Smart Citation

idtracker.ai: Tracking all individuals in large collectives of unmarked animals

Romero-Ferrero

Bergomi

Hinz

et al. 2018

Preprint

View full text Add to dashboard Cite

Our understanding of collective animal behavior is limited by our ability to track each of the individuals. We describe an algorithm and software, idtracker.ai, that extracts from video all trajectories with correct identities at a high accuracy for collectives of up to 100 individuals. It uses two deep networks, one detecting when animals touch or cross and another one for animal identi cation, trained adaptively to conditions and di culty of the video.Obtaining animal trajectories from a video faces the problem of how to track animals with correct identities after they touch, cross or they are occluded by environmental features. To bypass this problem, we proposed in idTracker the idea of tracking by identi cation of each individual using a set of reference images obtained from the video [1]. idTracker and further developments in animal identi cation algorithms [2-6] can work for small groups of 2-15 individuals. In larger groups, they only work for particular videos with few animal crossings [7] or with few crossings of particular species-speci c features [5].Here we present idtracker.ai, a system to track all individuals in small or large collectives (up to 100 individuals) at a high identi cation accuracy, often of > 99.9%. The method is species-agnostic and we have tested it in small and large collectives of zebra sh, Danio rerio and ies, Drosophila melanogaster. Code, quickstart guide and data used are provided (see Methods), and Supplementary Text describes algorithms and gives pseudocode. A graphical user interface walks users through tracking, exploration and validation (Fig. 1a).Similar to idTracker [1], but with di erent algorithms, idtracker.ai identi es animals using their visual features. In idtracker.ai, animal identi cation is done adapting deep learning [8][9][10] to work in videos of animal collectives thanks to speci c training protocols. In brief, it consists of a series of processing steps summarized in Fig. 1b. After image preprocessing, the rst deep network nds when animals are touching or crossing. Then the system uses the images between these detected to train a second deep network for animal identi cation. The system rst assumes that a single portion of video when animals do not touch or cross has enough images to properly train the identi cation network (Protocol 1). However, animals touch or cross often and this portion is then typically very short, making the system estimate that identi cation quality is too low. If this happens, two extra 1 . CC-BY-NC 4.0 International license not peer-reviewed) is the author/funder. It is made available under a

show abstract

Section: D43 Protocol 1: One-global-fragment Trackingmentioning

confidence: 99%

“…If more than 90% of the images in the global fragments are accumulated during the accumulation, we proceed to the identi cation of the individual fragments not used for training section D.5. Otherwise, we will repeat the accumulation starting from G σ (2) . If the accumulation fails even in this case, we repeat it with G σ(3) as a basis.…”

Section: Activity Mapmentioning

confidence: 99%

idtracker.ai: Tracking all individuals in large collectives of unmarked animals

Romero-Ferrero

Bergomi

Hinz

et al. 2018

Preprint

View full text Add to dashboard Cite

show abstract

“…Dolado et al . () developed an image‐processing method for the detection of occlusion in video and separate individuals involved in two‐dimensional (2D) space. Moreover, Baum et al .…”

Section: Preharvestmentioning

confidence: 99%

“…It can also be defined as the phenomenon of two or more tracked target objects becoming one during a time period (Delcourt et al 2012). Typically, there are two types of occlusion: (i) when two fish are swimming so close that can be falsely perceived as a single longer fish; and (ii) when two trajectories cross and two fish are perceived as a single T-or X-shaped individual (Dolado et al 2014). Detecting occlusion and tracking individuals during and after occlusion have been studied by some researchers.…”

Section: Developed a Computer Visionmentioning

confidence: 99%

“…P erez-Escudero et al (2014) expanded an algorithm by identifying individual fish, using fingerprints from occlusion-free portions of video and then used the signature to resolve the occlusions and identify switches. Dolado et al (2014) developed an image-processing method for the detection of occlusion in video and separate individuals involved in two-dimensional (2D) space. Moreover, Baum et al (2014) created an algorithm for overcoming occlusion problem by employing hybrid filter and game theory-related reinforcement.…”

Section: Developed a Computer Visionmentioning

confidence: 99%

See 1 more Smart Citation

Application of machine vision systems in aquaculture with emphasis on fish: state‐of‐the‐art and key issues

Saberioon

Gholizadeh

Císař

et al. 2016

Reviews in Aquaculture

185

109

View full text Add to dashboard Cite

Demands of aquatic products are increasing dramatically during past decades. Also quality assurance has gradually received more attention by both producers and consumers. Thus, fish producers are exploring all possible approaches for improving the productivity and profitability. Monitoring of fish state and behaviour during cultivation may help to improve profitability for producers and also reduce the threat of severe loss because of disease and stress incidents. It is necessary to evaluate and measure quality of fish products in accurate, fast and objective way for meeting the different demands of the fish‐processing industry after harvesting. Traditional methods are usually time‐consuming, expensive, laborious and invasive. Using rapid, inexpensive and noninvasive methods is therefore important and desirable. Optical sensors and machine vision system provide the possibility of developing faster, cheaper and noninvasive methods for in situ and after harvesting monitoring of quality in aquaculture. This review describes the most recent technologies and the suitability of different optical sensors for the fish farming management and also assessment, measurement and prediction of fish products quality. Two major areas of optical sensors applications in aquaculture are discussed in this review: (i) preharvesting and during cultivation; and (ii) post‐harvesting. Finally, accuracy and uncertainty of optical sensors applications in aquaculture are discussed. This review showed that MVSs and optical sensors have found real‐world application based on tremendous possibility offered by digital camera development and increasing the speed of computer‐based processing; however, still new algorithms, methods and re‐engineered sensors need to be developed to meet real‐world requirements.

show abstract

Zebrafish

Rihel¹,

Ghosh²

2016

Drug Discovery and Evaluation: Pharmacological Assays

View full text Add to dashboard Cite

A method for resolving occlusions when multitracking individuals in a shoal

Cited by 19 publications

References 24 publications

idtracker.ai: Tracking all individuals in large collectives of unmarked animals

idtracker.ai: Tracking all individuals in large collectives of unmarked animals

Application of machine vision systems in aquaculture with emphasis on fish: state‐of‐the‐art and key issues

Zebrafish

Contact Info

Product

Resources

About