Individual IDs

767 775 771 768

Prizes we want to opt for

Top 8 Prizes Most Entertaining Hack Most Annoying Hack Most Socially Useful Hack Best Hardware Hack


We came across hardware and technologies that we thought were interesting, and wanted to leverage on these state-of-the-art developments to put together something cool.

What it does

It is a self-driving RC car that is capable of following people autonomously. It tracks people using deep learning models specifically trained to identify people.

How we built it

We retrofitted a toy RC car from scratch with microcontrollers for low-level control, a NVIDIA Jetson TX2 for high-level logic and to support deep learning capabilities, and a ZED stereo camera for the perception of the environment.

The latest version of the Robot Operating System (ROS) was employed for high-level control, coupled with the most recent iteration YOLO (a convolutional neural network framework for object detection and classification) and statistical kernel-based tracking techniques.

Challenges we ran into

The lithium ion battery that we selected was not robust enough to support a surge in power output. This led to the disruption of continuous power supply to the various components, hence causing abrupt power shortage to various components under the presence of sudden changes in load. The solution was to run the power output from the battery through a power regulator unit to stabilize power to the various components.

Accomplishments that we're proud of

We were able to produce a robust product that proved to perform up to expectations. We were able to seamlessly integrate several recent technologies that were new to us within a short amount of time.

What we learned

We took a leap of faith to put this together within 24 hours, which seemed daunting to us at first. We believe that it was effective teamwork and believing in ourselves that were key in helping us materialize this complex idea. We learning that innovation and development at its core is about having vision and setting ourselves diligently toward the task.

What's next for FollowMeSempai

We plan to advance our detection, tracking, navigation and interaction capabilities by integrating various sensory fusion techniques, such relying on more accurate distance information produced by lidars or using encoders and IMU to provide more accurate odometry.

Built With

Share this project: