Farhan

DINOv3_Distillation_YOLO-pose. This project provides a complete pipeline to pre-train the backbone of a custom YOLOv11 pose estimation model using knowledge distillation from a powerful DINOv3 vision foundation model.

IntegraPose. IntegraPose, an open-source toolkit for training single, end-to-end models that perform simultaneous behavioral classification and keypoint estimation. IntegraPose leverages YOLO-Pose architectures and transfer learning, enabling researchers to create custom models via a streamlined annotation workflow.

DeepLabCut-Analysis-Jupyter-Scripts. While DLCAnalyzer and DLC Helper functions are great. They fail to accommodate a vast range of behavior investigator's needs. The Jupyter script provided here is meant for post processing/analyzing DeepLabCut (DLC) CSV files. The script allows user to filter data, calculate velocities, draw Region of Interest (ROI), and more.

Dynamic_Position_Velocity_Visualizer_For_DLC. I created this notebook to help me with behavioral neuroscience experiments. It calculates the average positions and velocities of two body parts (for better accuracy), and creates visualizations such as GIFs and streamline plots to represent the motion and flow of movement.

Gait_Analysis_DeepLabCut. Configurable Python project for analyzing gait in animal video experiments.

BehaviorAnalysis_YOLOv8. Python

Temporal_Behavior_Analysis. Temporal analysis tools.

DABA-Dynamic_Animal_Behavior_Analysis. Python

Multi-ROI-Analysis-with-DeepLabCut-CSV-Outputs. Jupyter Notebook

BehaviorDEPOT_Visualizations. This Repo contains Matlab scripts to aid in visualizing data ouput from BehaviorDEPOT.

ConvertFrames2OneHotFormat. I extract frames from a video using FFmpeg and identify interesting behaviors frame by frame. I then use a Jupyter script to parse out the frame numbers and create an annotation file, a CSV file, which can then be imported into our A-SOID project.

MitoFlow. 4D-mitochondrial analysis platform.

Gait_Analysis_YOLO. This repo is extension of DeepLabCut-based Gait analysis, but has been switched over now to use YOLO-focused pose structure. It ingests per-frame Ultralytics pose labels, converts them into a canonical dataframe, and then runs gait, pose, ROI, and Kuramoto-style limb coordination analysis on top of that data.