Automated Recognition of Social Behavior in Rats: The Role of Feature Quality

Lorbach, Malte; Poppe, Ronald; Dam, Elsbeth A. van; Noldus, L.P.J.J.; Veltkamp, Remco C.

doi:10.1007/978-3-319-23234-8_52

Cited by 9 publications

(11 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, the quality of tracking algorithms has a large impact on assessment accuracy (Lorbach et al, 2015). Visual object tracking is a fundamental and crucial problem in machine vision with several realworld applications (Smeulders et al, 2013;Yilmaz et al, 2006).…”

Section: Introductionmentioning

confidence: 99%

Markerless mouse tracking for social experiments

Sterley

Cheng

et al. 2021

Preprint

View full text Add to dashboard Cite

Automated behavior quantification requires accurate tracking of animals. Simultaneous tracking of multiple animals, particularly those lacking visual identifiers, is particularly challenging. Here we propose a markerless video-based tool to simultaneously track two socially interacting mice of the same color. It incorporates conventional handcrafted tracking and deep learning based techniques, which are trained on a small number of labeled images from a very basic, uncluttered experimental setup. The output consists of body masks and coordinates of the snout and tail-base for each mouse. The method was tested on a series of cross-setup videos recorded under commonly used experimental conditions including bedding in the cage and fiberoptic or headstage implants on the mice. Results obtained without any human intervention showed the effectiveness of the proposed approach, evidenced by a near elimination of identities switches and a 10% improvement in tracking accuracy. This suggests that the hybrid approach could be valuable for studying group behaviors, such as social interaction. This novel approach addresses problems of mistaken identities and lost information on key anatomical features that are common in existing methods. Finally, we demonstrated an application of this approach in studies of social behaviour of mice, by using it to quantify and compare interactions between pairs of mice in which some are anosmic, i.e. unable to smell. Our results indicated loss of olfaction impaired typical snout-directed social recognition behaviors of mice, while non-snout-directed social behaviours were enhanced.

show abstract

Section: Introductionmentioning

confidence: 99%

Markerless mouse tracking for social experiments

Sterley

Cheng

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…To classify rodent interactions, we use a log-linear classification model. The model can be learned efficiently, has only one free parameter and has performed sufficiently well in previous rodent behavior experiments [28,29]. The model allows for the calculation of a confidence score by using a suitable loss function, as we will explain shortly.…”

Section: Fig 2 Framework Components See Text For Detailsmentioning

confidence: 99%

“…Circling occurs only sporadically and is therefore a minority class in CRIM13. As in RatSI, we treat all close-contact interactions, which we cannot distinguish reliably from the trajectory-based features [29], as one contact behavior. Similarly, we join actions performed by the individual animal such as drink and eat and create a solitary class.…”

Section: Crim13 and Young Rats Datasetmentioning

confidence: 99%

Interactive rodent behavior annotation in video using active learning

Lorbach

Poppe

Veltkamp

2019

Multimed Tools Appl

Self Cite

View full text Add to dashboard Cite

Manual annotation of rodent behaviors in video is time-consuming. By learning a classifier, we can automate the labeling process. Still, this strategy requires a sufficient number of labeled examples. Moreover, we need to train new classifiers when there is a change in the set of behaviors that we consider or in the manifestation of these behaviors in video. Consequently, there is a need for an efficient way to annotate rodent behaviors. In this paper we introduce a framework for interactive behavior annotation in video based on active learning. By putting a human in the loop, we alternate between learning and labeling. We apply the framework to three rodent behavior datasets and show that we can train accurate behavior classifiers with a strongly reduced number of labeled samples. We confirm the efficacy of the tool in a user study demonstrating that interactive annotation facilitates efficient, high-quality behavior measurements in practice.

show abstract

“…Given the time and accuracy limitations of manual annotation, increasing work has focused on creating methods to automate the annotation process. Many such methods rely on tracking the animals’ bodies 4,6–9 or body parts 10 , from which higher-level features (e.g., velocity, acceleration, and posture) are extrapolated and used to classify behavior. Jhuang, et al 7 , for example, used motion and trajectory features to train a hidden Markov support vector machine to categorize eight classes of mouse behavior.…”

Section: Introductionmentioning

confidence: 99%

DeepAction: A MATLAB toolbox for automated classification of animal behavior in video

Harris

Finn

Tse

2022

Preprint

View full text Add to dashboard Cite

The identification of behavior in video is a critical but time-consuming component in many areas of animal behavior research. Here, we introduce DeepAction, a deep learning-based toolbox for automatically annotating animal behavior in video. Our approach uses features extracted from raw video frames by a pretrained convolutional neural network to train a recurrent neural network classifier. We evaluate the classifier on two benchmark rodent datasets and show that it achieves high accuracy, requires little training data, and surpasses both human agreement and similar existing methods. We also create a confidence score for classifier output, and show our method provides an accurate estimate of classifier performance and reduces the time required by human annotators to review and correct automatically-produced annotations. We release our system and accompanying annotation interface as an adaptable, non-technical, and open-source MATLAB toolbox.

show abstract

Automated Recognition of Social Behavior in Rats: The Role of Feature Quality

Cited by 9 publications

References 10 publications

Markerless mouse tracking for social experiments

Markerless mouse tracking for social experiments

Interactive rodent behavior annotation in video using active learning

DeepAction: A MATLAB toolbox for automated classification of animal behavior in video

Contact Info

Product

Resources

About