Passenger Safety in cab

women safety in cab around the world

Comment

Inspiration

What it does

A Real-Time Monitoring and Reporting of In-Cabin situations using Al, Computer Vision and Audio Analysis.

Detect events like driver swaps while ride, weapons in car, verbal and physical harassment, fights, etc.

A Dashboard for 24x7 incident response team to continuously monitor cabs.

Real-Time Video Feed starts if something goes wrong as judged by the algorithm.

Response time reduced and help can be sent immediately. Additional features like relatives overwatch and SMS alerts.

How I built it

PSC brings safety to the computer with with the help of intelligent in cabin monitoring done by psc running on driver mobile , psc video feed is continuously sent to the cloud where an AI algorithms calculates the risk and share live video feed when the risk is high . , To predict the risk we have categorized various sub problems into various problems ,

The first problem was Unauthorized driver for we have used HarrCascade and Open Cv.

Next problem was overall environment of the cabin For this, we created an emotion recognition model that uses facial features of the driver and passengers to predict the mode inside the cabin. Since computer vision algorithms tend to fail in dark. We also created an audio based emotion recognition model poses RNN model is used to isolate the frequency of human voice from the background noise, the audio emotion classifier operates on this isolated frequency to predict the mood of the conversation, namely angry scared neutral, etc. all the algorithms, were tested on various videos scraped from YouTube as well as real life scenarios in active bias.

The next problem for the detection of potential weapons inside the gap. For this, we used a custom yolo model trained with the coco data set. This model is used to detect harmful objects like scissors, knives, etc. to implement the machine learning models in real life, we migrated the detection algorithms to virtual instances on the cloud. Since processing power of smartphones very only a minimal model is run locally and the heavy models are run on the cloud. The dashboard communicates with the VM instance in real time to get all the parameters, while ensuring low latency, the safe cab app is a lightweight app that runs locally on the drivers smartphone. The app continuously sets the given footage to the virtual instance on the cloud. This has been facilitated by our very own compression algorithm that reduces the size of the video by almost 80%. The app also has internet less routing capabilities, safe care provides a dashboard for the gap aggregators like Ola and Uber, whoever 24 seven Incident Response Team. The dashboard reinforces the system and helps in avoiding false positives. It shows various gaps running within the cities and their parameters. It also shows the calves, who go off the grid during a trip with then went on to test our system in an actual gap in one

sense the camera module is just an ordinary smartphone app, it can easily be integrated with existing data services, interacting in real life scenario where the driver decides to leave the car, and hand it over to another unknown, the algorithms, the status of each car and also metadata associated with each. As we can see here, PSC is online and the risks are changing dynamically. Let's have a look what happens when Gabby is threatened by his driver, the person sitting at the dashboard can see what is happening inside the car, in real time, and report, in case of a threat, the system notifies the emergency contacts, the safe cab app allows the relatives to get the current location and the feed of the relatives in case of a threat.

Let's see what happens when a cab loses internet connectivity, a system is also immune to the case of internet loss. For this we are taking advantage of the internet, independent via p2p technology. This allows our application to quickly find and interact with nearby devices without needing to connect to a network or hotspot. Before any connection can be formed p2p devices find each other through listening and sending probe requests, with additional p2p. info on social channels in the scenario of a cab losing its internet, it can connect to nearby cabs through via p2p which then further relays the information to its nearby cabs until a cab with Internet is formed, which then sends the info to the cloud.

Get another extreme cases when the phone is switched off to tackle that we are working with beacon technology which is a device operated using Bluetooth Low Energy protocol which transmits radio signal and small packets of data to mobile devices. When a mobile comes in proximity range of a beacon. They send message to that device. So, when the phone is switched off in a cab. The other nearby cabs can detect the proximity, and ultimately the location of the cab,

Technologies used

To predict the risk we have categorized various sub problems into various problems ,

Unauthorized driver for that we have used HarrCascade and Open Cv. Overall environment of the cabin For this, we created an emotion recognition model that uses facial features of the driver and passengers to predict the mode inside the cabin. Detection of potential weapons inside the Cab.

Share this project:

Updates