Code of the Unsupervised Traffic Accident Detection paper in Pytorch.

No description available

Real-Time Accident Detection in Traffic Surveillance Using Deep Learning

IDL course project: Traffic Accident Detection via Deep Learning.

Accident detection system for traffic footage. Using computer vision and ML to detect and analyze accidents in a CCTV footage in real-time.

We proposed a large-scale benchmark for traffic accidents detection from video surveillance

No description available

STTAD is a dataset we propose for road traffic accident detection from surveillance views, featuring fine-grained classification and spatiotemporal annotations. This repository provides examples from the training and testing sets to showcase the data structure. The full dataset and related algorithms will be gradually updated.

Accident Detection Using YOLOv8 real-time object detection to identify traffic accidents from video feeds or footage.

While driving a car, collision detection plays an important role in autonomous driving and for the driver. This project presents an integrative approach to speed and direction detection which aims to be as simple as possible to enable the real- time computation while being able to adapt to a variety of highway and urban scenarios on different conditions. In this work, we implement a signal-based forward collision warning system and an overtaking vehicle detection system. The forward collision warning system aims to detect other vehicles driving ahead of the host vehicle while overtaking so as to avoid the potential traffic accidents. On the other hand, the overtaking vehicle detection system is to detect other vehicles behind or on the side of the host vehicle so as to give an alert to the driver when the vehicle begins overtaking. Experimental results are presented with real scene. The performance evaluation has demonstrated the effectiveness of the proposed techniques.

Project for paper "A Simulation-Based Framework for Urban Traffic Accident Detection in Surveillance Videos"

A cloud-based, AI-powered solution for real-time traffic monitoring and management. This system analyzes CCTV and GPS data to predict traffic density, prioritize emergency vehicles, and optimize traffic signal timing. Features include dynamic route suggestions, accident detection alerts, and a centralized dashboard for city-wide traffic control.

An accident detection model based on CNN(Convolutional Neural Network) for traffic cameras

Detects anomalies in traffic flow. If there is a green light and a traffic jam at the same time, it is an anomaly. The database gets updated and suitable alerts are given. The accident detection algorithm detects accidents in the anomaly.

Due to the growing population and people's need for comfort, more automobiles are being purchased, particularly in urban areas. This can result in heavy traffic, indicating that traffic violations are becoming more dangerous in every corner of the world. As a result, people's awareness decreases, and there are more accidents, which may result in the loss of many lives. The existing system has less accuracy and slow detection of violations, here we are using YOLO and OCR algorithms for object and number plate detection, these algorithms can detect the violation at high speed with good accuracy. The proposed system can detect the most common types of traffic violations in real-time through computer vision techniques and it also leverages good results with an accuracy of 88.3%. The proposed traffic violation detector can identify signal violations, and the individuals are informed that they will be apprehended if they break a traffic law. The proposed system is faster and more efficient than human, as known already traffic police is the one who captures the image of individuals violating traffic rules but the traffic police will not be able to capture more than one violation simultaneously. When compared to other algorithms YOLO is found to be more advantageous and has higher efficiency and accuracy.

Object identification has many applications in various fields like autonomous vehicles. In all over the world, important information about the road condition and its limitations are introduced to drivers as visual signals, such as traffic signs. Traffic signs are an important part of road infrastructure to provide information about the condition of road, warnings, prohibition, restriction, and other helpful information to the driver for navigation. They provide important information which can be interpreted by drivers. During inferior traffic or bad weather conditions, driver may not notice the signs directly or indirectly, which may lead to accidents or serious injuries. During such circumstances, if there is an automatic detection system for traffic signs, it can warn driver of such signs on the road and help him follow such signs and thus making driving safe. Advanced driver assistance system (ADAS) is one of the fastest growing fields in autonomous vehicle. ADAS technology is completely based upon vision system, active sensor technology and car data network. A vision-based road sign detection system is thus necessary to catch the driver’s attention to avoid any accidents. However, there are many factors which make the road sign detection difficult such as lighting conditions poor or bright, deformation of signs, angle at which they are placed. Thus, our aim in this project is to write an algorithm for vision-based traffic sign Identification.

Road traffic crashes are the leading cause of death among young people between 10 and 24 years old. In recent years, both academia and industry have been devoted towards the development of Driver Assistance Systems (DAS) and Autonomous Vehicles (AV) to decrease the number of road accidents. Detection of the road surface is a key capability for both path planning and object detection on Autonomous Vehicles. Current road datasets and benchmarks only depict European and North American scenarios, while emerging countries have higher projected consumer acceptance of AV and DAS technologies. This paper presents a selected Brazilian urban scenario dataset and road detection benchmark consisting of annotated RADAR, LIDAR and camera data. It also proposes a novel evaluation metric based on the intersection of polygons. The main goal of this manuscript is to provide challenging scenarios for road detection algorithm evaluation and the resulting dataset is publicly available at www.lrm.icmc.usp.br/dataset.

paper_list_traffic_accident_detection [WIP]

The Road Traffic Accident Analysis project explores patterns and causes of accidents using Python. It covers driver demographics, weather, vehicle conditions, accident causes, and time trends, using EDA, statistics, outlier detection, and ML to uncover insights for improving road safety.

Traffic Guardian is an innovative solution designed to enhance road safety and efficiency across California's busy highways. As a COS301 Capstone Project at the University of Pretoria, our goal is to develop a real-time traffic incident detection system that leverages existing traffic camera feeds to identify accidents, congestion and road hazards.

In developing nations such as India, the vehicular growth rate is increasing exponentially which is worsening the traffic operations. Most of the urban cities in India are facing traffic related problems such as congestion, accidents, pollution, etc. during peak hours. The main cause for traffic congestion in such cities is mainly due to uncontrolled urbanization and extensive usage of private vehicles. The traffic congestion leads to many problems like increase in travelling time, health disorders and accidents. Road accidents in India claimed over 1.5 lakh lives in the country in the year 2018, with over-speeding of vehicles being the major cause. The Ministry of Road Transport and Highways report on Road accidents in India stated that road accidents increased by a rate of 0.46 % in the year 2018 when compared to 2017 . Due to this there is a need to develop a model which can analyze and detect poor road conditions like potholes.This project aims in building a system which can detect the poor road conditions and can notify the driver as well as the government beforehand to improvise the road conditions. Pothole detection is being carried out using two techniques namely image processing and machine learning techniques. Those two techniques are used for a study of the detection and occurrence of potholes. In this project, we implemented both of them individually and then a combination of the techniques to see how image pre-processing can affect the performance of a deep learning model. The image pre-processing steps like erosion,median blur etc applied in this project removes the noise in the image which helps in better training of the model.First of all, we implemented the image processing techniques on a single image in the order: median blur, erosion, canny edge detection, contour detection, bounding box prediction. After that we labelled a dataset of around 800 images and passed it to the YOLOv5 model and noted the results.Secondly, we applied median blur on the already labelled dataset and then passed it to the YOLOv5 model and noted the results. Lastly, we applied median blur and erosion both and passed it to the model and noted the results. We compared the results at last. The conclusions reached are that a combination of machine learning and image processing techniques generates good performance in pothole detection and machine learning techniques provide better results than the usual image processing models.

An Intelligent Robotics Device (IRD) is developed in order to support safe mobility for disabled or elderly people with various impairments. The proposed system offers five main functions: obstacle detection and avoidance through bone conduction, live tracking, geofencing, GPS navigation. First, through a combination of an ultrasonic sensor and bone conduction phone, it detects diverse obstacles and produces that describe environmental information, including the positions and sizes of obstacles, which is then given to the learning-based algorithm. By learning the common patterns among assigned to the same directions, the IRD can automatically find paths to prevent collisions with obstacles. Second, it distinguishes a situation whereby the user is standing on a sidewalk, traffic intersection, or roadway through analysing the texture and shape of the images, which aids in preventing any accidents that would result in fatal injuries to the user, such as collisions with vehicles. In this project, a novel IRD system is presented that provides more safety for people with impairments and for elderly people. In order to assure safe mobility, our product offers gps live tracking, GPS navigation and geofencing. With these three functions, it can perceive obstacles of various types and recognize dangerous situations, and then recommend viable paths to evade them. First, obstacles are identified using a combination of ultrasonic ones, and then the avoidable directions are determined using learning based algorithms. Second, in order to prevent collisions with vehicles at traffic intersections, the situation recognition component distinguishes the type of place where the user is currently located as a sidewalk, an intersection. This work aims to provide safe mobility to wheelchair users while they control the wheelchair toward a destination. In order to support safe mobility, the wheelchair must detect a range of obstacles and dangerous situations in real environments and generate avoidable paths to prevent collisions with them. In order to achieve this, a hybrid obstacle avoidance method and a situation recognition method are proposed.

traffic accident detection using yolo

##//This Repository is about detection and alerting the driver when he is feeling drowsy Drowsy driving is one of the major causes behind fatal road accidents. One of the recent study shows that one out of five road accidents are caused by drowsy driving which is roughly around 21% of road accidents, and this percentage is increasing every year as per global status report on road safety 2015, based on the data from 180 different countries. This certainly highlights the fact that across the world the total numbers of road traffic deaths are very high due to driver’s drowsiness. Driver fatigue, drink-and drive and carelessness are coming forward as major reasons behind such road accidents. Many lives and families are getting affected due to this across various countries. All this led to the development of Intelligent Driver Assistance Systems. Real time drowsy driving detection is one of the best possible major that can be implemented to assist drivers to make them aware of drowsy driving conditions. Such driver behavioral state detection system can help in catching the driver drowsy conditions early and can possibly avoid mishaps. Among these the major cause is due to driver errors and recklessness. Driver fatigue is cause behind such mishaps. Heavy traffic, increasing automotive population, adverse driving conditions, tight commute time requirements and the work loads are few major reasons behind such fatigue. With this project, we are presenting technique to detect driver drowsiness using of Open CV and image processing.//##

Due to the growing population and people's need for comfort, more automobiles are being purchased, particularly in urban areas. This can result in heavy traffic, indicating that traffic violations are becoming more dangerous in every corner of the world. As a result, people's awareness decreases, and there are more accidents, which may result in the loss of many lives. The existing system has less accuracy and slow detection of violations, here we are using YOLO and OCR algorithms for object and number plate detection, these algorithms can detect the violation at high speed with good accuracy. The proposed system can detect the most common types of traffic violations in real-time through computer vision techniques and it also leverages good results with an accuracy of 88.3%. The proposed traffic violation detector can identify signal violations, and the individuals are informed that they will be apprehended if they break a traffic law. The proposed system is faster and more efficient than human, as known already traffic police is the one who captures the image of individuals violating traffic rules but the traffic police will not be able to capture more than one violation simultaneously. When compared to other algorithms YOLO is found to be more advantageous and has higher efficiency and accuracy.

No description available

Accurate accident detection and efficient traffic control are crucial for improving road safety. Deep learning techniques have demonstrated remarkable potential in this domain. This repository focuses on utilizing deep learning models to detect accidents and control traffic, leveraging large-scale datasets for training and optimization.

Advanced neural network algorithms are employed for the identification and classification of different vehicle categories on a large set of traffic surveillance data.Implemented hashing algorithm for eliminating vehicle duplication issue, thereby optimizing the speed of detection of vehicles. This traffic analysis thereby reduces traffic congestion and accidents and further helps in smart city implementation.

Kapaan is an AI-driven system for real-time accident detection, emergency alerts, and traffic signal automation. Built for the UYIR Road Safety Hackathon 2025, it won Best Performing Team & ₹10,000. Developed using AI, FastAPI, Flutter & Firebase, it ensures faster response & safer roads.

The development of the self-driving car is needed for the safety of driver and passenger on the vehicle. The two most important parts of advanced driver assistance systems are a collision avoidance system and a Lane keeping assist system, which could help to reduce the number of traffic accidents. A fundamental technique for effective collision avoidance and lane-keeping is a robust lane detection method