The problem DriveSharp today for a possible tomorrow solves

The danger of accidents brought on by driver exhaustion or sleepiness can be significantly reduced by driver alertness, which is a crucial component of road safety. A buzz sound produced by our Android app can significantly increase driver awareness and encourage safer driving conditions.

The buzz sound feature can be triggered by the app when it detects signs of sleepiness or sleep deprivation in the driver. The sound can be a loud, attention-grabbing alarm that alerts the driver to the fact that they are in a dangerous state and need to pull over and take a break. The sound can be customized to be as loud as necessary to ensure that the driver is alerted and takes action.

Reduced Accidents: The app can assist in reducing accidents brought on by driver inattentiveness or sleep deprivation by monitoring the frequency of yawning and blinking. If the app indicates that the driver is in a risky situation, the emergency SOS feature can notify the driver or a selected emergency contact, lowering the chance of traffic accidents.

Challenges we ran into

1. Integration of external camera with the Android Studio : After the latest versions of Android 11, there are no pre-made API’s or frameworks which allow us to capture videos from an external camera and store it in an android device. It is due to the recent privacy concern arising in the androids and the support system is getting steadily depreciated. For this, we tried to use Raspberry Pi microcontroller as it provides a suitable platform for integrating the camera module. Due to being light-weight, it helps in ensuring real-time deployment of the project which ensures that the driver’s video feed is continuously processed and classified if the driver is drowsy or not.
2. Number Details pane : We also ran into the problem of storing user provided number for SOS correspondence so that the messages can be sent to the prescribed users. This was due to the fact that there were errors while using shared resources in the Java files. you will need to initialize Shared Preference instance inside the pre-defined class.
3. Deployment of the ML model inside Android studio : We were also facing some issues while deploying the pre-trained Deep Learning model when we were integrating with Java source files. As we tried to integrate the tflite model in Java, we found that it was depriceated and hence hard to deploy. For this, we tried to utilize the Chaquopy SDK in Python for android, but it was not returning as expected.