Object Detection in Autonomous Vehicles

Introduction

In this project, our goal is to write software that can detect humans, vehicle, roadlane from an autonomous vehicle's perspective.

Result

Human Detection:

Roadlane Detection:

Vehicle Detection:

Team

Group: 22
Members: Sriram Akella, Hana Myre, Martin Mollestad, Morteza Haidari,

Project Structure

.
├── .gitignore
├── data
│  ├── CarPOV_h-detect.mp4
│  ├── CarPOV_roadlane.mp4
│  ├── CarPOV_v-detect.mp4
│  ├── coco.names
│  ├── extra
│  ├── yolov3-320.cfg
│  └── yolov3-320.weights
├── notes.md
├── output
│  ├── v-detect
│  └── vehicle_detection_v3.mp4
├── ReadMe.md
└── src
   ├── demo.sh
   ├── extra
   ├── human-detect.py
   ├── humanDetact.py
   ├── roadlane.py
   ├── tracker.py
   └── vehicle_detect.py

Folders:

Src - Folder where our source code is located
- extra - This folder contains code that were unnecessary for final result, but were used for early versions.
Data - This folder contains our datasets, such as videos and images.
- extra - This folder contains datasets that were unnecessary for final result, but were used for early versions.
Output - This directory contains output files from code. NOTE: This folder contains files that were not used in the final version.

Files:

The follow files were used to achieve project goals:

Human Detections:

human-detect.py

Roadlane Detections:

roadlane.py

Vehicle Detections:

System Design

Prerequisite

- Python>=3.8
- imutils==0.5.4
- matplotlib==3.4.3
   - cycler [required: >=0.10, installed: 0.11.0]
   - kiwisolver [required: >=1.0.1, installed: 1.3.2]
   - numpy [required: >=1.16, installed: 1.21.4]
   - pillow [required: >=6.2.0, installed: 8.4.0]
   - pyparsing [required: >=2.2.1, installed: 3.0.5]
   - python-dateutil [required: >=2.7, installed: 2.8.2]
      - six [required: >=1.5, installed: 1.16.0]
- opencv-python==4.5.4.58
   - numpy [required: >=1.17.3, installed: 1.21.4]
- scipy==1.7.2
   - numpy [required: >=1.16.5,<1.23.0, installed: 1.21.4]

Usage

cd src
bash demo.sh or ./demo.sh

FAQ

Why is the repo called hbd?

Originally, we thought of human detection, i.e human body detection = hbd. After our first presentation, we moved to Object detection in autonomous vehicles.

Srm-Akla/ODAV