Pytorch, OpenCV, C++, DNN, Python
Overview: Running on little Rpi4, this AI-powered algorithm can detect if a driver is sleepy, distracted, smoking, drinking or yawning while driving. Intended to be used by fleet owners to monitor their drivers and prevent accidents.
Metrics: 15 FPS, 0.97 ACC, 0.02 FP, 0.01 FN
Weakness: Head down, low light, sunglasses, mask

Deepstream-6.0, C++, OpenCV, Gstreamer, DeepSort, Python, Pytorch, ONNX
Overview: Running on Jetson Nano, this custom AI model was trained on Visdrone dataset. The live inference was made to demonstrate the use of AI-powered drones for traffic monitoring.
Metrics: 40 FPS, 0.91 ACC

OpenCV, C++
Overview: Drone is equipped with a sensor that detects gas leakage (green laser inside circle). This laser pointer has to be on the pipeline to work accurately. The non-AI algorithm uses Lucas-Kanade, Match Template, and noise reduction algorithms to detect and track pipelines.
Then it calculates the angle that needs to be given to the gimbal to keep laser pointer on the pipeline.
Intended to assist drone pilots so that they don't have to correct gimbal frequently.
Weakness: objects overlapping pipeline

OpenCV, Python, Numpy
Overview: The main frame demonstrates the position of drone, with grids representing 1x1 meter in real world. Top right and bottom right frames are simulating the object's position and drone's altitude.
The algorithm then calculates the relative angle and relative speed that needs to be given to the drone to follow the object.
Weakness: Frequent loss of detection

Deepstream-6.0, C++, OpenCV, Gstreamer, DeepSort, Python, Pytorch, ONNX
Overview: Running on Jetson Nano, this custom deepstream-app communicates with our dashboard to draw region of interests, crossing lines, etc.. to analyze traffic situation.
Intended to be used by traffic control to assist in development of smart city.

OpenAI gpt-3.5, MySQL, PHP, JS, EC2, Route53, Cloudfront
Overview: Launched this website way before chatGPT, before all the Generative AI hype. It generates code, linux and windows commands, SQL Queries, and live CSS based on given prompt.
Weakness: chatGPT

Tensorflow, OpenCV, Python
Overview: Trained an AI model to detect asteroids in telescopic (.fits) images distributed by IASC. The model detected three Asteroids out of which two were officially given preliminary status.
Weakness: short asteroid streak, asteroid overlapping another star

Python, Flask, Google Vision, OpenAI gpt-3.5
Overview: Upload any image and it'll detect the main object, primary and secondary colors, generate tagline, description, primary and secondary categories.
Developed it for a client that onboards and manage products for online sellers.
Weakness: Blurry image, too many products in a single image, etc.