DeepBhvTracking

DeepBhvTracking performs automated animal behavior detection and tracking using the You Only Look Once (YOLO) algorithm and background subtraction to produce bounding boxes and centroids for movement quantification in neuroscience experiments.

Key Features:

• Integration of Deep Learning: Utilizes the You Only Look Once (YOLO) algorithm combined with background subtraction to detect and localize animals in video frames.
• Customizable Training: Trains the detector with manually labeled images and a pretrained deep-learning neural network to adapt across experimental conditions and animal models.
• Robust Tracking Capability: Generates bounding boxes and tracks target centroids using background subtraction, maintaining performance under uneven illumination and recording-device interference.

Scientific Applications:

• In Vivo Optical Imaging and Electrophysiological Recording: Enables precise movement tracking during in vivo optical imaging and electrophysiological recording to support analysis of neural circuit function.
• Behavioral Paradigms: Applies to diverse behavioral paradigms for studies of movement disorders, social deficits, and mental disease-related behaviors.

Methodology:

The detector is trained on manually labeled images with a pretrained deep-learning neural network integrated with the YOLO algorithm; the trained detector generates bounding boxes around target animals in footage; tracking computes target centroids within bounding boxes using background subtraction.