Accident-Detection

Overview

Our main goal of this project is to use deep learning and computer vision to detect accidents on dashcam and report it to nearby emergency services with valid accident images.

Challenges

1. Our main challenge was to gather accident images and videos and manually categuorize images into accient and non-accident frames

2. To design a deep convolutional neural networks model for this project.

3. Limited hardware resorces like GPU's.

Team Members

1: Manoj Pawar Sj Github LinedLn

2: Manjunath Inti Github LinedLn

3.Bharat Kaushik Github LinedLn

4.DIkshita Basu Github LinkedLn

5.Bikash Singha Github LinkedLn

6.Rema Rose Toppo Github LinkedLn

Model Overview

1 . For this project we have tweaked Densenet-161 architecture

Dense Convolutional Network (DenseNet), connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections - one between each layer and its subsequent layer - our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters.

The 1-crop error rates on the imagenet dataset with the pretrained model are listed below.

Model structure Top-1 error Top-5 error

densenet121 : 25.35 : 7.83

densenet169 : 24.00 : 7.00

densenet201 : 22.80 : 6.43

densenet161 : 22.35 : 6.20

Prerequisite

Download anaconda from here https://www.anaconda.com/distribution/#download-section

1. Pytorch

conda install pytorch torchvision cudatoolkit=10.1 -c pytorch

2. OpenCV

conda install -c conda-forge opencv

3. Dataset of accident/non-accident images

https://drive.google.com/open?id=1o0D7vnGUZHS72is6n1jV1ge2BDfObzVi

4. Pretrained Model binary file

https://drive.google.com/open?id=1AnJSogx65iyfIG0cSm5D15xfTGJzst8d

5. A proper php-language environment like xampp,remove htdocs folder and replace that with htdocs in this repo

6. Clone or Download this repo

git clone https://github.com/manojpawarsj12/accident-detection.git

DEMO

1.accident

2.Non-accident

Train

Go to bash/cmd and type

python train.py

Tensorboard visual

1. Traning set2. Validation set3.Number of corercts v/s epochs

Test/Accuracy

Go to bash/cmd and type

python test.py

Test on video

python evaluate.py

Test on Webcam

python livewebcam.py

Result

The model reaches a classification accuracy of 86.00% accuracy on a randomly sampled test set, composed of 20% of the total amount of video sequences from our dataset. Will re-train this model when we have a good GPU and somre data .