DeePosit, an AI-based tool for detecting mouse urine and fecal depositions from thermal video clips of behavioral experiments

Scientists conduct behavioral experiments on animals to study brain mechanisms that govern social behavior and how these may be affected by various conditions. For example, in rodents, urination and defecation are important social activities used for communication and territory marking, and they are influenced by the emotional state of an individual.

In the past, these activities were analyzed at the end of an experiment by shining ultraviolet light on a filter paper placed on the floor of the cages. However, this method does not provide information on when urination or defecation occurred. Also, in many cases, urine drops are smeared on the filter paper due to the animal's movement during the experiment, which reduces the accuracy of this method. To bridge this gap, Peles et al. developed a computer-vision algorithm – named DeePosit – to automatically track mice's urination and defecation activities during social behavior experiments recorded with a thermal camera.

To examine the efficiency of the tool, the researchers analyzed the urination and defecation activities of mice during several social behavior tests. They then tested whether these activities changed over time and if there were differences between male and female mice, or between different strains of laboratory mice.

The analysis revealed that the tool could identify the time and location of each urination and defecation event with an accuracy similar to that of a human observer. Using this tool, Peles et al. demonstrated that urination and defecation activities changed during a social encounter, for example, urination became more frequent. They observed that males urinated more often than females, which may be attributed to differences in their territorial behavior. It also revealed differences between laboratory strains.

Peles et al. are confident that this rapid, unbiased and cost-effective tool can improve the analysis of social behavior in animals, particularly rodents. This will be especially relevant for researchers investigating the effect of treatments in mouse models of various disorders. The tool can also be trained and adapted to different behavioral and experimental contexts. It may allow a comparison of an additional important aspect of social behavior in treated and non-treated animals, and in health and disease.

In many mammalian species, including rodents, social interactions are accompanied or followed by events of active urination, also known as micturition or voiding activity (Arakawa et al., 2008). Multiple studies have demonstrated that urine and fecal deposits comprise many chemosensory social signals that carry information about the individual, such as its species, sex, social rank, and identity, as well as its reproductive and health conditions (Bigiani et al., 2005). These chemosensory signals include various metabolites, as well as many proteins, such as major urinary proteins (Brennan, 2004). Thus, by depositing urine spots and feces in its environment, the individual also deposits social information, which may later be perceived by other individuals and modify their future social interactions with this individual (Hurst and Beynon, 2004). In other words, the use of urine and fecal deposits allows individuals to advertise their availability to possible mates and communicate with other conspecifics. Moreover, in territorial species, urination is used to mark the territory of the individual, thus functioning as a spatio-social scent-marking activity (Brennan and Kendrick, 2006). In rodents, urination was shown to be strongly influenced by the individual’s internal state, social rank, social context, and previous social experience (Desjardins et al., 1973; Hyun et al., 2021). Therefore, monitoring urination activity can provide valuable information on the individual’s social behavior and internal state. Specifically, deficits in urine depositing may reflect atypical social behavior in rodent models of various diseases (see Wöhr et al., 2011 for example), hence may be used for testing potential treatments in such models.

Urination during a given task is traditionally analyzed via the void spot assay, which uses filter paper placed on the arena floor to analyze, after the end of the experiment, the spatial distribution of urine spots (Wolff and Powell, 1984; Higuchi and Arakawa, 2022). However, this analysis usually lacks the temporal dimension, is distorted by urine smearing across the arena floor caused by the individual’s movement (see Figure 2d, e), and is limited in detecting overlapping urine spots. Another caveat is that the filter paper may be torn down by the mouse during the behavioral experiment. Recently, Dalghi et al., 2023 used a filter paper on the arena floor, UV light, several cameras, and a manual video annotation to analyze urination events. Several other studies (Verstegen et al., 2020; Miller et al., 2023a) used thermal imaging via infrared (IR) camera for such analysis, as urine deposits are emitted while being in body temperature, hence can be seen in the thermal image. However, fecal deposits are also emitted in body temperature, making it difficult to distinguish between feces and small urine spots by thermal imaging alone. Moreover, these studies relied on manual analysis of thermal video clips, which made the analysis process time-consuming and subjected to observer bias. To cope with these limitations, we have developed an open-source computer vision-based software to automatically detect and classify deposited urine and feces from thermal video clips. Our detection and classification algorithm is based on a combination of a heuristic algorithm used for the preliminary detection of bright (warm) blobs in the thermal video clip and a trainable video classifier used to classify the preliminary detections as either urine, feces, or background (BG, i.e., not urine or feces). We demonstrate the efficiency of this tool by analyzing the temporal dynamics of urination and defecation activities in male and female CD1 (ICR) mice while performing three social behavior tests, and further validate the algorithm by testing it with male C57BL/6J mice. We found that urination and defecation activities show distinct dynamics across the various tests in a sex-, strain-, and test-dependent manner.

Here, we present a new algorithm and an open-code trainable AI-based computational tool for detecting and classifying urination and defecation events from thermal video clips. This algorithm enables a detailed characterization of the dynamics of urination and defecation activities during social behavior of small rodents. One advantage of this tool is that it is automated, thus allowing a rapid and observer-unbiased analysis of urine and fecal deposition events and areas, with a good temporal and spatial resolution. Specifically, combining our algorithm with an IR camera for thermal imaging of behavioral experiments can replace the void spot test, which usually lacks any temporal resolution and is prone to mistakes caused by urine smearing and filter-paper tearing. Finally, our algorithm facilitates the analysis of defecation activity, which was rather unexplored so far but may contribute to scent-marking behavior, as discussed below. Our algorithm uses thermal video clips generated by an IR camera placed above the arena and does not require a camera placed below a clear arena floor, as used by a recent paper (see Keller et al., 2018 for example). Thus, it can be utilized for analyzing experiments conducted in standard experimental setups, such as those used for the three-chamber test. The computational tool and experimental method presented here may be useful for a detailed characterization of social behavior in mice, including murine models of autism spectrum disorder and other pathological conditions. It may also be used to explore urination and defecation activities in other scientific contexts, unrelated to social behavior. Finally, our experimental setup is cheap and easy to assemble, and the detection algorithm can run on a standard PC with a GPU card.

Analysis of the errors made by the algorithm in the test dataset (see Figure 2—video 1 for video clips of these events) raised several limitations, that might be addressed in future work. Urine or fecal deposits must be fully visible while the deposit is still warm. A close adjacency between the mouse and the deposit might cause the mouse mask to overlap the mask of the deposit, thus preventing its detection. Many of the ‘miss’ events in the test video clips were created by the mouse staying close to the urine or fecal deposits for a long period after their deposition. Few other ‘miss’ events were due to very small urine spots or due to repeated urination in the same position during a very short time period, which resulted in detecting these separate urination events as a single event by the algorithm. A wrong classification of urine as fecal deposition occurred in 2.3% of the urination events. In many of these events, the urination spot was small (and therefore harder to distinguish from a fecal deposition) (see Figure 3—figure supplement 3c). Wrong classification of background as feces occurred 21 times in the test set. In most of these events, the mistake was due to feces that were moved by the mouse to a new location while still being warm. Such cases may be mitigated in future work by a tracking algorithm that continuously tracks the location of each fecal deposit. Wrong classification of background as urine occurred 33 times in the test set, with some of these errors caused by smearing of large warm urine spots.

We evaluated the accuracy of the algorithm and found it to be uniform across the various sexes, tests and session stages of the experiments used by us. This suggests that the low level of errors made by the algorithm should not create a bias during biological experiments. Moreover, the algorithm achieved a good and stable accuracy even for C57BL/6 mice examined in a while arena, a condition that was not represented in the training videos. Thus, the algorithm seems to be robust, with a low sensitivity to changing conditions. We also compared the algorithm’s accuracy to the accuracy achieved by a second human annotator on the same dataset and concluded that the algorithm accuracy is comparable to the accuracy of a human annotator while being much faster and unbiased. Finally, the algorithm showed superior performance over classic object detection algorithms, such as YOLOv8, which are based on a single image input. This is most likely due to the transformer-based architecture of our algorithm, which allows it to use the temporal information extracted from the thermal video clips.

Future work might improve DeePosit by extending the training set and including more challenging examples. Notably, comparing a small training set (Figure 3h) with a larger one (Figure 2h) shows that the larger training set improved the accuracy of DeePosit. Another way for future improvement in DeePosit accuracy may be by using a trainable detection and segmentation algorithm instead of heuristic preliminary detection. Note that our classifier currently does not get the mask of the preliminary detection as an input, making the classification task harder when there are adjacent deposition events. An end-to-end trainable detection, segmentation, and classification pipeline might address these limitations but will require a much larger training set. Future work might also adapt the algorithm for multi-animal experiments. Such adaptation might require detecting the mask of each of the animals, separating the identity of each of the animals, and associating each deposition with the relevant animal.

We validated our method and algorithm using experimental results from social discrimination tests conducted by male and female CD1 and male C57BL/6 mice. We demonstrated distinct dynamics of urination and defecation activities across the habituation and trial stages, with sex-, test-, and strain-dependent differences. Both male and female CD1 mice, as well as male C57BL/6 mice showed higher rates of defecation activity at the early stage of the habituation phase, as compared to later stages (Figure 5). This tendency may reflect a higher level of anxiety at the beginning of the habituation phase, caused by the novel context. Still, it may also serve for scent-marking activity, that labels the arena as a familiar environment. The latter explanation is supported by the fact that the peak in defecation activity was not reduced from the first-day test (SP) to the second and third-day tests (SxP and ESPs), when the subject is expected to be less anxious due to the familiar context. In contrast to defecation, urination activity at the beginning of the habitation phase in CD1 mice was test dependent. While no peak was observed during the SP test, the first time the animals were exposed to the experimental arena, it was observed in the second test (SxP) and got even stronger in the last test (ESPs). This development was statistically significant in CD1 males but not in females. Since these changes occur during the habituation phase, before the introduction of stimuli to the arena, they cannot reflect the type of test and thus seem to be induced by the order of the experiments. Notably, similar dynamics across experimental days were previously reported using the void spot assay for C57BL/6j mice (Keil et al., 2016). This suggests that the induction of urination activity by males at the early stage of the habituation phase represents territorial scent-marking activity, which is positively correlated with the higher familiarity experienced by the subject in the arena as the experiments progressed between days. It should be noted that an early peak of urination upon entering an environment was reported by a recent study using a thermal camera for manual analysis of urination activity (Miller et al., 2023b). A second peak of urination activity was observed at the beginning of the trial period, after stimuli insertion to the arena. This was observed in both male and female CD1 mice, but the test type significantly affected it only in males. In this case, we cannot dissect the effect of test type from the test order, as the urination activity occurred after stimuli insertion and, hence, may be induced by the presence of specific social stimuli. Since the subjects are already habituated to the arena at this stage, the elevated urination activity seems to serve as part of the subjects’ social behavior, most probably as a territorial scent-marking behavior induced by the presence of social stimuli, i.e., competitors. Interestingly, we found several differences in the dynamics of CD1 male mice and C57BL/6 male mice, suggesting that the scent-marking behavior is also strain specific. Unlike CD1 male mice, C57BL/6 male mice exhibited a peak in urination already at the beginning of the first (SP) habituation, a trend toward higher level of defecation activity in the SP trial stage, and no increase in urination activity during the SP and SxP trial stage, compared to the habituation end. However, several findings were common for both CD1 and C57BL/6 male mice, such as the higher feces rate at the beginning of habituation in comparison to the end of habituation and the higher levels of urination at the beginning of the SxP habituation stage.

We did not observe a consistent spatial distribution of the urine or fecal deposits between the arena sides of the preferred and non-preferred stimuli in CD1 mice. This seems to contradict a recent study (Miller et al., 2023b), that reported opposite bias toward familiar versus unfamiliar stimuli in losers versus winners wild-derived mice following a social contest. This contradiction may be due to the distinct mouse strains or the distinct contexts of social behavior (presentation of a single stimulus animal in comparison to two simultaneously presented animals) used by both studies.

Overall, the novel algorithm and software presented here enable a cost-effective, rapid, and unbiased analysis of urination and defecation activities of behaving mice from thermal video clips. The algorithm is trainable and may be adapted to various behavioral and experimental contexts. Thus, it may pave the way for the integration of this important behavioral aspect in the analysis of small rodents’ social and non-social behaviors, in health and disease.

Code for computing Two Way Chi-Square Test which was used to compare the distribution of active mice (with at least one detection) in males vs females.

% Compute two way chi square test for 2x2 table
%
% Hypothseis H0: there is no relation between gender and the distribution of zeros .
% Hypothseis H1: there is a relation between gender and the distribution of zeros .
%
% Inputs :
% valsMales - a vector of length 2 that contains : [zeros count , non zeros count] for males .
% valsFemales - a vector of length 2 that contains : [zeros count , non zeros count] for females.
%
% Outputs :
% pVal - p value . A value lower than 0.05 suggests that the hypothesis H0 should be rejected .
% chiStat - statistic of the chi square test .
% df - degree of freedom (equals 1 for 2x2 tables).
%
function [pVal , chiStat ,df] = TwoWayChiSqrTest (valsMales , valsFemales)
if length (valsMales)~=2 || length (valsFemales)~=2
‍‎error (’input vectors should have length =2 ’)
end
sumMales = sum (valsMales);
sumFemales = sum (valsFemales);
sumAll = sumMales + sumFemales ;
sum1 = valsMales (1) + valsFemales (1) ;
sum2 = valsMales (2) + valsFemales (2) ;
expectedFreqMales = [sumMales *(sum1 / sumAll), sumMales *(sum2 / sumAll)];
expectedFreqFemales = [sumFemales *(sum1 / sumAll), sumFemales *(sum2 / sumAll)];
chiStatMales = sum ((valsMales - expectedFreqMales).^2 ./ expectedFreqMales);
chiStatFemales = sum ((valsFemales - expectedFreqFemales).^2 ./ expectedFreqFemales);
chiStat = chiStatMales + chiStatFemales ;
df = 1;
pVal = 1- chi2cdf (chiStat ,df);

Raw video files will be shared with any academic PI upon a reasonable request from the corresponding author (davidpelz@gmail.com) and within a reasonable time. The data is free to use for academic purposes. No project proposal is required. For commercial use, please contact Shlomo Wagner (shlomow@research.haifa.ac.il). The code, trained classifier weights, statistical data, and an example of a raw video are shared in GitHub (copy archived at Peles, 2025) and Zenodo.

1. 1. David P
   2. Shai S
   3. Natalie R
   4. Taghreed S
   5. Shlomo W

   (2025) Zenodo

   DeePosit: an AI-based tool for detecting mouse urine and fecal depositions from thermal video clips of behavioral experiments.

   https://doi.org/10.5281/zenodo.14754159

1. 1. Arakawa H
   2. Blanchard DC
   3. Arakawa K
   4. Dunlap C
   5. Blanchard RJ

   (2008) Scent marking behavior as an odorant communication in mice

   Neuroscience and Biobehavioral Reviews 32:1236–1248.

   https://doi.org/10.1016/j.neubiorev.2008.05.012

   • PubMed
   • Google Scholar
2. 1. Benjamini Y
   2. Hochberg Y

   (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing

   Journal of the Royal Statistical Society Series B 57:289–300.

   https://doi.org/10.1111/j.2517-6161.1995.tb02031.x

   • Google Scholar
3. Book

   1. Bigiani A
   2. Mucignat-Caretta C
   3. Montani G
   4. Tirindelli R

   (2005) Pheromone Reception Inmammals Berlin

   Springer.

   https://doi.org/10.1007/s10254-004-0038-0

   • Google Scholar
4. 1. Brennan PA

   (2004) The nose knows who’s who: chemosensory individuality and mate recognition in mice

   Hormones and Behavior 46:231–240.

   https://doi.org/10.1016/j.yhbeh.2004.01.010

   • Google Scholar
5. 1. Brennan PA
   2. Kendrick KM

   (2006) Mammalian social odours: attraction and individual recognition

   Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 361:2061–2078.

   https://doi.org/10.1098/rstb.2006.1931

   • PubMed
   • Google Scholar
6. Conference

   1. Carion N
   2. Massa F
   3. Synnaeve G
   4. Usunier N

   (2020) End-to-end object detection with transformers

   European Conference on Computer Vision. pp. 213–229.

   https://doi.org/10.1007/978-3-030-58452-8_13

   • Google Scholar
7. Conference

   1. Cartucho J
   2. Ventura R
   3. Veloso M

   (2018) Robust object recognition through symbiotic deep learning in mobile robots

   2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). ages.

   https://doi.org/10.1109/IROS.2018.8594067

   • Google Scholar
8. 1. Dalghi MG
   2. Montalbetti N
   3. Wheeler TB
   4. Apodaca G
   5. Carattino MD

   (2023) Real-time void spot assay

   Journal of Visualized Experiments 01:e64621.

   https://doi.org/10.3791/64621

   • PubMed
   • Google Scholar
9. 1. Desjardins C
   2. Maruniak JA
   3. Bronson FH

   (1973) Social rank in house mice: differentiation revealed by ultraviolet visualization of urinary marking patterns

   Science 182:939–941.

   https://doi.org/10.1126/science.182.4115.939

   • Google Scholar
10. Conference

    1. He K
    2. Zhang X
    3. Ren S
    4. Sun J

    (2016) Deep residual learning for image recognition

    2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). pp. 770–778.

    https://doi.org/10.1109/CVPR.2016.90

    • Google Scholar
11. 1. Higuchi Y
    2. Arakawa H

    (2022) Contrasting central and systemic effects of arginine-vasopressin on urinary marking behavior as a social signal in male mice

    Hormones and Behavior 141:105128.

    https://doi.org/10.1016/j.yhbeh.2022.105128

    • PubMed
    • Google Scholar
12. 1. Hurst JL
    2. Beynon RJ

    (2004) Scent wars: the chemobiology of competitive signalling in mice

    BioEssays 26:1288–1298.

    https://doi.org/10.1002/bies.20147

    • PubMed
    • Google Scholar
13. 1. Hyun M
    2. Taranda J
    3. Radeljic G
    4. Miner L
    5. Wang W
    6. Ochandarena N
    7. Huang KW
    8. Osten P
    9. Sabatini BL

    (2021) Social isolation uncovers a circuit underlying context-dependent territory-covering micturition

    PNAS 118:e2018078118.

    https://doi.org/10.1073/pnas.2018078118

    • PubMed
    • Google Scholar
14. Software

    1. Jocher G
    2. Chaurasia A
    3. Qiu J

    (2023) Ultralytics yolov8, version 8.3.159

    GitHub.

    https://github.com/ultralytics/ultralytics
15. 1. Keil KP
    2. Abler LL
    3. Altmann HM
    4. Bushman W
    5. Marker PC
    6. Li L
    7. Ricke WA
    8. Bjorling DE
    9. Vezina CM

    (2016) Influence of animal husbandry practices on void spot assay outcomes in C57BL/6J male mice

    Neurourology and Urodynamics 35:192–198.

    https://doi.org/10.1002/nau.22692

    • PubMed
    • Google Scholar
16. 1. Keller JA
    2. Chen J
    3. Simpson S
    4. Wang EH-J
    5. Lilascharoen V
    6. George O
    7. Lim BK
    8. Stowers L

    (2018) Voluntary urination control by brainstem neurons that relax the urethral sphincter

    Nature Neuroscience 21:1229–1238.

    https://doi.org/10.1038/s41593-018-0204-3

    • PubMed
    • Google Scholar
17. 1. Kopachev N
    2. Netser S
    3. Wagner S

    (2022) Sex-dependent features of social behavior differ between distinct laboratory mouse strains and their mixed offspring

    iScience 25:103735.

    https://doi.org/10.1016/j.isci.2022.103735

    • PubMed
    • Google Scholar
18. 1. Mervis CB
    2. Dida J
    3. Lam E
    4. Crawford-Zelli NA
    5. Young EJ
    6. Henderson DR
    7. Onay T
    8. Morris CA
    9. Woodruff-Borden J
    10. Yeomans J
    11. Osborne LR

    (2012) Duplication of GTF2I results in separation anxiety in mice and humans

    American Journal of Human Genetics 90:1064–1070.

    https://doi.org/10.1016/j.ajhg.2012.04.012

    • PubMed
    • Google Scholar
19. 1. Miller CH
    2. Haxhillari K
    3. Hillock MF
    4. Reichard TM
    5. Sheehan MJ

    (2023a) Scent mark signal investment predicts fight dynamics in house mice

    Proceedings. Biological Sciences 290:20222489.

    https://doi.org/10.1098/rspb.2022.2489

    • PubMed
    • Google Scholar
20. 1. Miller CH
    2. Hillock MF
    3. Yang J
    4. Carlson-Clarke B
    5. Haxhillari K
    6. Lee AY
    7. Warden MR
    8. Sheehan MJ

    (2023b) Dynamic changes to signal allocation rules in response to variable social environments in house mice

    Communications Biology 6:297.

    https://doi.org/10.1038/s42003-023-04672-x

    • PubMed
    • Google Scholar
21. 1. Mohapatra AN
    2. Peles D
    3. Netser S
    4. Wagner S

    (2024) Synchronized LFP rhythmicity in the social brain reflects the context of social encounters

    Communications Biology 7:2.

    https://doi.org/10.1038/s42003-023-05728-8

    • PubMed
    • Google Scholar
22. 1. Netser S
    2. Haskal S
    3. Magalnik H
    4. Wagner S

    (2017) A novel system for tracking social preference dynamics in mice reveals sex- and strain-specific characteristics

    Molecular Autism 8:53.

    https://doi.org/10.1186/s13229-017-0169-1

    • PubMed
    • Google Scholar
23. 1. Netser S
    2. Haskal S
    3. Magalnik H
    4. Bizer A
    5. Wagner S

    (2019) A system for tracking the dynamics of social preference behavior in small rodents

    Journal of Visualized Experiments 01:e60336.

    https://doi.org/10.3791/60336-v

    • Google Scholar
24. Software

    1. Netser S

    (2020) TrackRodent, version 7d91b09

    GitHub.

    https://github.com/shainetser/TrackRodent
25. Software

    1. Peles D

    (2025) DeePosit, version swh:1:rev:be73ff51549f95ad7cc4b67dfc497ba03650a8d2

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:2f68a84e482cac6aa72dde94e47fa3a07615faf0;origin=https://github.com/davidpl2/DeePosit;visit=swh:1:snp:0ecb1949aed9d5608e1fdc2669a89188e9f226a2;anchor=swh:1:rev:be73ff51549f95ad7cc4b67dfc497ba03650a8d2
26. 1. Verstegen AM
    2. Tish MM
    3. Szczepanik LP
    4. Zeidel ML
    5. Geerling JC

    (2020) Micturition video thermography in awake, behaving mice

    Journal of Neuroscience Methods 331:108449.

    https://doi.org/10.1016/j.jneumeth.2019.108449

    • PubMed
    • Google Scholar
27. 1. Wegner KA
    2. Abler LL
    3. Oakes SR
    4. Mehta GS
    5. Ritter KE
    6. Hill WG
    7. Zwaans BM
    8. Lamb LE
    9. Wang Z
    10. Bjorling DE
    11. Ricke WA
    12. Macoska J
    13. Marker PC
    14. Southard-Smith EM
    15. Eliceiri KW
    16. Vezina CM

    (2018) Void spot assay procedural optimization and software for rapid and objective quantification of rodent voiding function, including overlapping urine spots

    American Journal of Physiology. Renal Physiology 315:F1067–F1080.

    https://doi.org/10.1152/ajprenal.00245.2018

    • PubMed
    • Google Scholar
28. 1. Wöhr M
    2. Roullet FI
    3. Hung AY
    4. Sheng M
    5. Crawley JN

    (2011) Communication impairments in mice lacking Shank1: reduced levels of ultrasonic vocalizations and scent marking behavior

    PLOS ONE 6:e20631.

    https://doi.org/10.1371/journal.pone.0020631

    • PubMed
    • Google Scholar
29. 1. Wolff PR
    2. Powell AJ

    (1984) Urine patterns in mice: An analysis of male/female counter-marking

    Animal Behaviour 32:1185–1191.

    https://doi.org/10.1016/S0003-3472(84)80235-3

    • Google Scholar

Author details

1. David Peles

   Sagol Department of Neurobiology, Faculty of Natural Sciences, University of Haifa, Haifa, Israel

   Contribution

   Formal analysis, Writing – original draft, Algorithm development, Video annotation

   For correspondence

   davidpelz@gmail.com

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0009-0003-5543-1676

2. Shai Netser

   Sagol Department of Neurobiology, Faculty of Natural Sciences, University of Haifa, Haifa, Israel

   Contribution

   Writing – review and editing, Algorithm development, Experimental setup building

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-4176-1124

3. Natalie Ray

   Sagol Department of Neurobiology, Faculty of Natural Sciences, University of Haifa, Haifa, Israel

   Contribution

   Data curation, Performing social behaviour tests, Video annotation

   Competing interests

   No competing interests declared

4. Taghreed Suliman

   Sagol Department of Neurobiology, Faculty of Natural Sciences, University of Haifa, Haifa, Israel

   Contribution

   Data curation, Performing social behaviour tests, Video annotation

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0009-0009-4177-5391

5. Shlomo Wagner

   Sagol Department of Neurobiology, Faculty of Natural Sciences, University of Haifa, Haifa, Israel

   Contribution

   Resources, Funding acquisition, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7618-0752

• 1,145

  views

• 10

  downloads

• 0

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.