Monitoring poultry social dynamics using colored tags: Avian visual perception, behavioral effects, and artificial intelligence precision

Abstract

Artificial intelligence (AI) in animal behavior and welfare research is on the rise. AI can detect behaviors and localize animals in video recordings, thus it is a valuable tool for studying social dynamics. However, maintaining the identity of individuals over time, especially in homogeneous poultry flocks, remains challenging for algorithms. We propose using differentially colored "backpack" tags (black, gray, white, orange, red, purple, and green) detectable with computer vision (eg. YOLO) from top-view video recordings of pens. These tags can also accommodate sensors, such as accelerometers. In separate experiments, we aim to: (i) evaluate avian visual perception of the different colored tags; (ii) assess the potential impact of tag colors on social behavior; and (iii) test the ability of the YOLO model to accurately distinguish between different colored tags on Japanese quail in social group settings. First, the reflectance spectra of tags and feathers were measured. An avian visual model was applied to calculate the quantum catches for each spectrum. Green and purple tags showed significant chromatic contrast to the feather. Mostly tags presented greater luminance receptor stimulation than feathers. Birds wearing white, gray, purple, and green tags pecked significantly more at their own tags than those with black (control) tags. Additionally, fewer aggressive interactions were observed in groups with orange tags compared to groups with other colors, except for red. Next, heterogeneous groups of 5 birds with different color tags were videorecorded for 1 h. The precision and accuracy of YOLO to detect each color tag were assessed, yielding values of 95.9% and 97.3%, respectively, with most errors stemming from misclassifications between black and gray tags. Lastly using the YOLO output, we estimated each bird's average social distance, locomotion speed, and the percentage of time spent moving. No behavioral differences associated with tag color were detected. In conclusion, carefully selected colored backpack tags can be identified using AI models and can also hold other sensors, making them powerful tools for behavioral and welfare studies.

Keywords: Computer vision; Japanese quail; Poultry; Visual model; YOLO.

Fig. 1

Experimental protocol. (a-e) Top panels represent photographs of experimental equipment. (a) Pavo cristatus visual model adapted from (Hart, 2002; Vorobyev, 2003). (b) Example of black backpack tag. (c) Experimental pen. (d) Color tags. (e) Female Japanese quail with orange tag. (f-h) An illustrative image as well as a brief description of each experiment is provided. (f) Reflectance spectra of backpacks (solid line, with colors representing the color of the tag) and feathers fom different birds (discontinuous lines). A larger version of this panel is provided as Supplementary Fig. 1. (g) Experimental pen with three quails with an orange label. (h) Experimental pen with five quails with different colors tags (white, black, gray, red, and green).

Fig. 2

Chromatic (dS) and achromatic (dL) color perception contrasts between colored backpack tags and feathers. Feathers were obtained from 5 different female Japanese quail as indicated with the circle color. The red line in both panels indicates the threshold of 5 JND (Just Noticeable Differences) above which the bird is considered to perceive the contrast between the feather and tag.

Fig. 3

Pecking behavior performed within homogeneous colored tag groups. Number of (a) pecks performed towards their own accelerometer. (b) pecks performed towards the accelerometer of conspecifics, (c) pecks performed towards the head or body of conspecifics and (d) pecks received by conspecifics directed to the head or body. Each individual is represented with a circle, for boxplots, boxes represent 25th and 75th percentiles of the sample data, the middle red line the 50th, and the whiskers the 1.5 times the interquartile range. ^a-c Post hoc analysis results are shown using blue lower-case letters. Groups that do not share the same letter differ significantly (P<0.05).

Fig. 4

Detection capabilities of the computer vision model in correctly classifying individuals based on tag color in mult-colored tag social groups. (a) Confusion Matrix was estimated using RoboFlow. (b) Overall correct identification was estimated after running the model on 1 h long behavioral videos, with a 1 s sampling rate. The percentage of overall correct identifications was calculated based on the total time points a bird with a colored tag appeared in the corresponding frame. If the tag was found twice that time point was not counted as a correct identification. Boxplots are presented for each tag color, with each colored circle representing the overall average for each color tag across all 10 groups studied. All colors, except black and gray, showed over 90% correct identification (dotted black line) in all groups.

Fig. 5

Examples of behavioral estimations using spatial coordinates of birds with colored tags. (a) Image representing the social distance between each bird in a social group using colored lines at a specific time point. (b) Using the same color scheme as in panel a, the social distance between each individual is shown as a function of time. Thick, dotted black line shows the estimated social circle. (c) The distance ambulated by each individual during the 1 s intervals is shown as a function of time. (d) Heat map of the average social distance between each bird during the 1 h test. Reds indicate birds staying farther apart than the estimated social circle and in blue those closer than the social circle. (e–i) Heatmaps of spatial use of each bird.

Fig. 6

Dynamics within multi-colored tag social groups. (a) Mean speed and (c) percent of time spent ambulating by each bird within the ten social groups. No significant differences were found between color tags.