Automated image-based tracking and its application in ecology
read more
Citations
DeepLabCut: markerless pose estimation of user-defined body parts with deep learning
Using DeepLabCut for 3D markerless pose estimation across species and behaviors
Toward a Science of Computational Ethology
DeepPoseKit, a software toolkit for fast and robust animal pose estimation using deep learning
Applications of machine learning in animal behaviour studies
References
Multiple view geometry in computer vision
Multiple View Geometry in Computer Vision.
Observational study of behavior: sampling methods.
KinectFusion: Real-time dense surface mapping and tracking
Related Papers (5)
Frequently Asked Questions (16)
Q2. What are the advantages of unsupervised methods?
Unsupervised techniques offer the advantage of decreased subjectivity, and increased throughput, repeatability, and the chance of finding rare behaviors [68,74,75].
Q3. What is the common approach for detecting individuals?
(A–C) A common approach for detecting individuals is background subtraction, where detection of individuals in raw images is achieved by removing an estimated background-only image, resulting in isolation of foreground pixels.
Q4. What is the way to track a scene in 3D?
(F) Light-field cameras allow for post-hoc selection of focal points, thus potentially allowing tracking and construction of the scene in 3D from a single image point.
Q5. What are the methods for quantifying the physical structure of 3D landscapes?
Methods for quantifying the physical structure of 3D landscapes are rapidly advancing [58– 60] and can be used for rendering features of natural habitats, such as trees or streams.
Q6. What is the final step in automated tracking?
The final step in automated image-based tracking is analysis, where position and pose data are analyzed to understand relevant biological, and ecological, patterns and processes.
Q7. What are the main issues that need to be addressed?
In addition, the storage and management issues that arise from the huge amounts of digital data that are easily produced by imaging must be addressed.
Q8. What are some of the more applied questions that can be addressed by image-based tracking?
Image-based tracking can also address more applied questions, such as the role of fragmentation in population dynamics (A.I. Dell, unpublished) or determining the size of animal populations that are historically difficult to measure [52].
Q9. What is the way to maintain identity in 3D landscapes?
This often involves application of artificial markings; however, natural variation in the morphology of individuals can also be used to maintain identities throughout image sequences, even following occlusion (Table S1 in the supplementary material online).
Q10. What is the way to quantify the environment?
Remote quantification of the environment can easily be accomplished by imaging in the appropriate sensory regime, such as optical video cameras for quantifying light conditions and thermal cameras for quantifying the thermal landscapes.
Q11. What are the advantages of lightfield cameras?
Lightfield cameras work at higher frame rates and there are several laboratories exploring if they can be successfully incorporated into automated tracking systems (I.D. Couzin and G.G. de Polavieja, unpublished).
Q12. What is the basic question that can be addressed by a simple analysis?
Once coordinates (and pose estimates if available) are produced, then even very simple analysis can address basic ecological questions such as where and how animals behave and interact [4,8] (Figure IA–C).
Q13. What can be done to improve the accuracy of the 3D imaging?
As in 2D, multiple 3D imaging cameras can be employed simultaneously to provide additional resolution and to cope with occlusions [29].
Q14. How can the position and pose of organisms be determined?
The position and pose of organisms with stiff and simple-shaped bodies can be computed by fitting a shape contour to the image of the organism [8,27] (Figure ID),including determining whether clumps of pixels should be separated into multiple individuals (Figure IE–I).
Q15. What constraints on the acquisition, processing, and storage of digital information limit the spatial extent of image?
constraints on the acquisition, processing, and storage of digital information limit the spatiotemporal extent of image-based tracking, and extracting the position and pose of every individual in each image is difficult in complex habitat and at high densities.
Q16. What are the general traits that can be used to maintain identities in ecology?
General traits can be sufficient for maintaining identities at low densities or when individuals vary greatly in size or shape, but in many other instances in ecology individuals are likely to be similarly sized or shaped.