diff --git a/_portfolio/portfolio-2.html b/_portfolio/portfolio-2.html
deleted file mode 100644
index 8a0c6ebb444e1..0000000000000
--- a/_portfolio/portfolio-2.html
+++ /dev/null
@@ -1,7 +0,0 @@
----
-title: "Portfolio item number 2"
-excerpt: "Short description of portfolio item number 2
"
-collection: portfolio
----
-
-This is an item in your portfolio. It can be have images or nice text. If you name the file .md, it will be parsed as markdown. If you name the file .html, it will be parsed as HTML.
diff --git a/_portfolio/portfolio-2.md b/_portfolio/portfolio-2.md
index 7928931576c4a..e1e165583b30f 100644
--- a/_portfolio/portfolio-2.md
+++ b/_portfolio/portfolio-2.md
@@ -1,23 +1,18 @@
---
-title: "Modelling and Control of a Rotary Inverted Pendulum"
-excerpt: "PID, LQR, and energy-based swing-up LQR on Quanser Qube Servo 2.
"
+title: "Recognizing and Tracking Person of Interest: A Real-Time Efficient Deep Learning based Method for Quadcopters"
+excerpt: "Participated in MATLAB and Simulink Challenge Projects program. Extended to a conference paper."
+
"
collection: portfolio
---
-This project was done as part of the course ELEC-E8004 Project Work course at Aalto University. The primary objective of this project was to develop and implement updated lab assignments and pre assignments for the control labs in the ELEC-C1310: Laboratory Exercises in Automation and Control Engineering course at Aalto University. Leveraging the Quanser Qube Servo 2 system, the project aims to provide students with practical, hands-on experience in control. On top of the course content, the scope is extended to non-linear control, and an energy-based swing-up combined with an LQR for balancing has been implemented.
+The recognition and tracking of a person of interest is a crucial task in many applications, including search and rescue, security, and surveillance. This paper presents a distributed system architecture that leverages the asynchronous threading and communication property of ROS2 to develop and implement a real-time efficient Deep Learning (DL) based method for recognizing and tracking a person of interest. The DL model receives snapshots from the quadcopter's camera and sends back an information vector, which includes all recognized persons and their corresponding position information within the camera frame of the quadcopter. The person of interest tracking control system receives face set information about the person of interest and generates reference velocity signals to be tracked by low-level controllers embedded within the drone. Experiments conducted in a cluttered and complex environment demonstrate the efficiency of the DL-based architecture for quadcopters. The presented real- world results validate the effectiveness of the proposed approach in recognizing and tracking a person of interest.
## Video Demonstration
-Below is a video demonstrating the implementation of control methods on the Quanser Qube Servo 2:
+Below is a video demonstrating our work:
-[![Quanser Qube Servo 2 Video](https://img.youtube.com/vi/YevSQ600GKA/0.jpg)](https://www.youtube.com/watch?v=YevSQ600GKA)
+[![Real-time Demo](https://img.youtube.com/vi/i7bYXnRy8Vc/0.jpg)](https://www.youtube.com/watch?v=i7bYXnRy8Vc)
## Images
-
diff --git a/images/drone.jpg b/images/drone.jpg
new file mode 100644
index 0000000000000..be007ca2f6d59
Binary files /dev/null and b/images/drone.jpg differ