TRex

TRex performs automated markerless visual tracking of multiple animals in video to quantify positions, outlines, 2D visual fields, head and rear positions, and to support behavioral and sensory experiments including open- and closed-loop paradigms.

Key Features:

• High-Speed Tracking: Operates at up to 60 Hz and can track approximately 256 individuals simultaneously.
• Background Subtraction: Uses background-subtraction techniques to isolate individuals from complex backgrounds in real time.
• 2D Visual Field and Posture Estimation: Provides 2D estimates of visual fields and outlines, and detects head and rear positions of bilateral animals.
• Markerless Identification with Deep Learning: Applies deep-learning algorithms for markerless identification of up to approximately 100 unmarked individuals, with reported speeds 2.5–46.7× faster and memory usage 2–10× lower than comparable software as group size or video length increases.
• Long-Duration and High-Resolution Video Support: Handles long-duration and high-resolution video streams for extended continuous monitoring.
• Support for Open- and Closed-Loop Experiments: Enables analysis and real-time performance suitable for both open- and closed-loop experimental contexts.

Scientific Applications:

• Ethology and Behavioral Analysis: Quantifies positions, postures, and visual fields to study animal behavior and decision-making.
• Neuroscience: Provides precise estimates of visual input and head/rear orientations to support sensory and neural investigations.
• Ecology and Field Studies: Enables long-duration, high-resolution monitoring for ecological fieldwork and laboratory experiments.

Methodology:

Combines background-subtraction techniques with deep-learning-based algorithms for markerless identification and real-time tracking.