How projetcomputervision/computervision works
Project type and purpose
• This is a research-oriented computer vision pipeline written in Python for extracting and analyzing object trajectories from video, with steps for camera calibration, homography, distortion evaluation, detection, tracking, and trajectory prediction (including gravitational adjustment).
• Key indicators: file names like Calibration.py, Homographie.py, detection_Videos.py, kalman.py, trajectoire_prediction.py, and sample videos under Videos_Detection.
File structure
• Top level scripts:
• AjoutGravite.py – applies gravity correction to computed trajectories
• Calibration.py – performs camera calibration
• Homographie.py – computes homography transform
• EvaluationDisortion.py – measures and evaluates lens distortion
• detection_Videos.py, Videos_Detection_3.py, Videos_Homographie.py – video-based object detection
• kalman.py – implements a Kalman filter for tracking
• trajectoire_prediction.py – predicts future positions
• Trajectoire_v3.py / Trajectoire_v4.py – trajectory computation pipelines
• Data/ folder: stored calibration and homography data (homography_data.npz)
• Videos_Detection/ folder: sample input/output videos showing detection masks and tracking
• Scripts_Secours/: backup scripts for calibration
• README.md and TODO.md: high-level project notes and next steps
README
• Contains minimal info (“Computer Vision – Test du git”), so rely on file names to deduce architecture.