Einstein-Vision

Usage Instructions

1. Install all dependencies (core depdendencies listed below) - NOTE: please review script before running, depthpro in particular has a large install size Run ./install_dependencies.sh
2. Parse data into proper format. Using P3Data with (Assets, Calib, Sequences) subfolders: Run ./Code/parse_data.sh

Dependencies:

Depth Pro: https://github.com/apple/ml-depth-pro/ Lane Segmentation (only need checkpoint): Article, https://debuggercafe.com/lane-detection-using-mask-rcnn/ Download, https://drive.usercontent.google.com/download?id=1WRu0e5GYlsKmQp0UR_ZyuYoXQmQWR4Pj&export=download&authuser=0

HumanPose - "yolo11x-pose.pt"

Vehicle classification - https://github.com/MaryamBoneh/Vehicle-Detection yolov11 - https://docs.ultralytics.com/models/yolo11/#performance-metrics

Collison Prediction - https://github.com/Abdelrhman-Amr-98/Pedestrian-Tracking-and-Collision-Prediction/

Data (name P3Data, place in root): https://app.box.com/s/zjys9xcefyqfj2oxkwsm5irgz6g3hp1l

Github repo Traffic Color detection:https://github.com/bhaskrr/traffic-sign-detection-using-yolov11/blob/main/process_video.py Pretrained Model - https://github.com/bhaskrr/traffic-sign-detection-using-yolov11/tree/main/model

1. clone repository and follow instructions
2. run process image.py