Keep Me Safe

Saving Changing lives.

We are three high school students designing and building a device that alerts car owners when a child or pet is left behind in a hot car. We integrate existing off-the-shelf electronics, a suite of sensors and an Artificial Intelligence-based computer vision system to recognize when there is imminent danger for occupants

Extreme heat is often labelled the silent killer. Many children and pets are left behind or gain access to cars every year. The inside of a vehicle can get very hot within minutes, even when the outside temperature is moderate.

Prolonged exposure to extreme heat can cause heat stroke and lead to injury and death. Approximately 60% of deaths are due to children being left in cars by accident, 25% are due to children gaining access to a vehicle, and 15% are due to children being left in cars intentionally.

Several countries have enacted legislation requiring new cars to be fitted with sensors. Existing aftermarket products are inadequate to provide a comprehensive solution. At KeepMeSafe, we are developing a device that utilizes a range of sensors and Artificial Intelligence to recognize when a child or pet is in danger and alerts the vehicle's owner.

Existing aftermarket products mainly aim to prevent drivers from accidentally leaving a child behind. Our device tracks the temperature inside a vehicle, checks for movement using a PIR sensor if the temperature exceeds a threshold and utilizes a computer vision system to detect the presence of a sleeping child that a PIR sensor would not detect.

United Nations Economic Commission for Europe statement

No systematic collection of vehicular heatstroke fatalities in the Asian, Australian, European continents, neither by NGO, individual countries or the EU. Unknown how many fatalities/incidents really occur around the globe. It is also known that pets and elderly occupants can die in overheated vehicles

Gallery

All AI Box Design Detection Hardware Flowcharts

HARDWARE SELECTION

Two aluminium heat sink plates attached above and below the Raspberry Pi Zero 2’s processor with thermal paste help dissipate the heat generated when the AI computer vision model is running.

FLOWCHARTS

Description

HARDWARE SELECTION

The Raspberry Pi Pico W controls the PIR sensor, heat sensor, LED indicators, and transistor that controls the Pi Zero 2 W’s power.

DETECTING SLEEPING CHILDREN

The back-facing camera allows the AI computer vision system to detect the presence of children in the back seats even when they are motionless. A PIR sensor may not be able to detect a sleeping child if the child remains motionless.

HARDWARE SELECTION

The Raspberry Pi Zero 2 W runs the camera, controls the GPS and establishes GSM-based communication with the vehicle’s owner.

CAMERA

The camera provides the input for the AI computer vision model, which detects the presence of humans and pets in the vehicle's back seats with a high degree of accuracy. The camera uses a fisheye lens for 160° image capture.

PIR SENSOR

The PIR sensor captures the movement of warm objects in the front and back seats. The windows reflect the infrared wavelength, ensuring that there are no false positives from passersby.

SAFE PETS

Our sensor’s camera can recognize common pets like cats and dogs.

EASY TO USE & EASY TO CHARGE

The sensor will sit on the armrest between the front seats of a car. It can be charged via its USB Type C charge port and will incorporate a 5,000 mAh battery pack.

3D PRINTING

We are using a 3D printer to test prototype designs for our box.

TWO WAYS TO DETECT THOSE IN DANGER

While the PIR sensor can detect the motion of warm objects, such as a human or a pet, it may not detect a sleeping child if the child remains motionless. Our device incorporates a camera and an AI system to detect such cases.

COMPACT BOX DESIGN

Our box design ensures that all sensor components fit into a compact space.

HIGHLY ACCURATE DETECTION

The AI model can detect humans and dogs even at a high threshold for positive identification. In this picture, the system is being tested on a computer and laptop camera rather than the sensor’s Pi Zero 2 W and its camera.

TESTING THE AI SYSTEM

Here, we are testing the AI model for occupancy detection via a computer and laptop camera. The system can localize and identify multiple people.

Compact Design

We are utilizing 3D printing technology to build and test the prototype compact box to house our device.

View Video

Development Steps

White Paper

Our white paper delves into the crucial topics of how heat affects the human body physiologically, the thermodynamics of heat transfer, deaths and injuries from extreme heat in vehicles, and the growth of Internet of Things (IoT) devices, all of which are pivotal to the development of our heat detection and occupancy alert system.

Decision Tree

We devised a solution incorporating an in-vehicle device with GSM connectivity, a central server, and a mobile telephone application. We also outlined a decision tree for how the device will process inputs from its sensors and camera and alert the vehicle’s owner and, in extreme cases, the emergency services.

Hardware Selection

We agreed on a set of sensors, the camera and lens, the means of communication and the microprocessor and microcontroller required for our device. We used off-the-shelf components throughout our project. We chose Pi W and Zero 2W from the Raspberry family of single-board computers and peripherals compatible with them.

AI Model Integration

We successfully tested an AI-driven computer vision system that can detect children and pets in vehicles, a significant milestone in our project. We installed it on our microprocessor using Linux and Phyton. We used input from the Raspberry Pi Camera, which we fitted with a fisheye lens for a wider image capture angle.

Box Design

We are currently designing a box for our device, which will be placed on the armrest between the front seats. We are designing the box using Autodesk Fusion and 3D-printing its components. In addition to a functional prototype on a breadboard, our goal is to have the final version of our device installed and integrated within the box.

Software Integration

We are currently writing the program to run both the Raspberry Pi W and Pi Zero 2 W. We are using Microphyton to program the Pi W and Phyton to program the Zero 2 W. We are integrating the two via a Wi-Fi solution. The Zero 2 W controls the camera, GPS, and GSM, while the Pi W controls the PIR and heat sensors.

Testing

We tested each sensor, the AI computer vision system, GPS, and GSM connectivity separately. We also tested the camera and AI system inside a car to ensure that they could detect humans and pets even when they were motionless. Once software and hardware integration is completed, we will test our device in a real-world setting.

Complete Initial Phase

We intend to complete Phase 1 when the software integration for our device is complete, we successfully test our device in a real-world setting where it is able to detect an occupant both via the PIR sensor and the AI system and when our device is able to warn the vehicle’s owner via a one-way SMS message.

Next Phase

In Phase 2, we intend to incorporate server connectivity for our device. We will also develop a mobile application that will allow the user to set up an online account, connect a particular device to this account and update the settings of a device. This connectivity will allow the device to send push notifications to the vehicle’s owner.

Saving Changing lives.

United Nations Economic Commission for Europe statement

Gallery

HARDWARE SELECTION

FLOWCHARTS

HARDWARE SELECTION

DETECTING SLEEPING CHILDREN

HARDWARE SELECTION

CAMERA

PIR SENSOR

SAFE PETS

EASY TO USE & EASY TO CHARGE

3D PRINTING

TWO WAYS TO DETECT THOSE IN DANGER

COMPACT BOX DESIGN

HIGHLY ACCURATE DETECTION

TESTING THE AI SYSTEM

Compact Design

Development Steps

White Paper

Decision Tree

Hardware Selection

AI Model Integration

Box Design

Software Integration

Testing

Complete Initial Phase

Next Phase

38

average number of child deaths per year from extreme heat in cars in the U.S. alone

20

number of minutes before the inside of a car reaches 38° C when it is 20° outside

60

degrees in Celsius the inside of a car can reach in one hour

10

percentage of deaths from extreme heat in a car when the outside temperature was below 27° C

Our Team

Ceylin Asil

Melis Ozvardar

Miray Karsidag