Vitalii Lakusta’s Post

I’m excited to share my project PyTerminal, a fully functional command terminal built in Python for the #SRMHacksWithCodemate hackathon! PyTerminal mimics a real system terminal with support for file and directory operations, system monitoring, and AI-driven natural language commands. Key Features: Execute standard commands like ls, cd, pwd, mkdir, rm, and move; monitor CPU, memory, and running processes; undo/redo file operations; understand natural language commands (e.g., “create folder test and move file1.txt into it”); log session history; and enjoy command history with tab-completion. I relied solely on CodeMate Ai tools to develop and debug the code efficiently. #SRMhacksWithCodemat CodeMate® AI

🚀 Excited to share my latest project: Face and Hand Gesture Recognition! I’ve developed a real-time application using Python, OpenCV, and MediaPipe that can: Track facial landmarks 🧑🦱 Detect hand landmarks ✋ Recognize common hand gestures automatically with emojis: 👍 Like ✌️ Peace 👊 Fist ✋ Stop 👋 Hi / Bye 🤟 Love 💪 Super / Power 👌 OK 💡 This project can be extended for human-computer interaction, gesture-based controls, virtual reality, or sign language recognition. #Python #OpenCV #MediaPipe #ComputerVision #AI #GestureRecognition #MachineLearning #ProjectShowcase #Innovation

Project : Hand Gesture Controlled Hill Climbing Simulation I buit a real-time hill climbing simulation controlled entirely by hand gestures using Python, OpenCV, and CVZone. Process: 1) Captured hand gestures using a webcam and CVZone's HandDetector. 2)Mapped gestures to control the movement of the hill climbing algorithm. 3)Visualized the algorithm in real-time with OpenCV. 4)Integrated AI optimization to make gesture control intuitive and interactive. This project demonstrates the fusion of computer vision, AI optimization, and interactive algorithm visualization, making algorithms engaging and easy to understand. #python #pytonprojects #begginerprojects #OpenCV #ComputerVision #AI #MachineLearning #Python #HandTracking #GestureRecognition #Automation #PyAutoGUI #CVZone #ArtificialIntelligence #TechProjects #Innovation #CodingLife #AIProjects

Meta has introduced the Code World Model (CWM), a tool designed to bridge the gap between code generation and understanding. CWM acts like a 'neural debugger,' simulating program behavior and predicting outcomes such as infinite loops with high accuracy. The model was trained using over 120 million Python program executions and can predict algorithm complexity, ranking second on the BigOBench benchmark. Meta has released CWM as an open-weights model for research purposes, emphasizing its focus on programming and mathematical reasoning. Read more: https://lnkd.in/eyPKtYsK 📰 Subscribe to the weekly AI Programming Weekly: https://lnkd.in/eUQ-KTc2 #ai #artificialintelligence #ainews

#SRMHacksWithCodemate CodeMate® AI Built an Advanced Python Terminal as part of my project! This terminal goes beyond traditional CLI by integrating AI-powered natural language commands, system monitoring, and file operations in a single interface. Features: Standard commands: ls, cd, mkdir, rm, pwd AI Natural Query: Run commands like “create a folder called demo” or “move file1.txt into demo” System monitoring: cpu, mem, ps Command history tracking I developed and executed this project in VS Code CLI, leveraging the Codemate Build extension for building and testing. This project shows how AI can enhance developer productivity by making terminals more intuitive and user-friendly compared to existing ones.

I ve always believed that coding is just a tool, while algorithms are the true heart of the IT industry. To reflect this, I ve been working on hand-written AI algorithms, which I ve documented and shared on my website. It s my way of exploring the foundations of intelligence in computing — not just using pre-built tools, but understanding the logic that powers them. You can check it out here: https://lnkd.in/d-EcjnPU #Python #ComputerVision #OpenCV #Mediapipe #Gaming #AI #Innovation #HandGestureRecognition #TechProjects #RealTimeControl #LearningByBuilding #MachineLearning #GitHub #OpenSource #Developers #Algorithms #DeepLearning #Transformers #neuralnetworks #MultiHeadSelfAttention #Embedding #PatternRecognition #ComputerVision

You often find inspiration in the most immediate problems. For me, it was the condition of the road right outside my hostel. That's what motivated me to build this AI-powered pothole detector.💭💭💭 Using Python, OpenCV, and a custom YOLO model—which I trained on Google Colab using their T4 GPU—the system processes video footage to identify and flag potholes. The video I used for testing is of that very road. To make this happen, I created my own dataset from scratch on Roboflow, meticulously drawing polygon segments to annotate each pothole. My ultimate goal was to use instance segmentation for precise edge detection—outlining the exact shape of the pothole rather than just boxing it in. While the dataset is capable of this, I ran into challenges deploying the segmentation🤷♂️🤷♂️ technique in the current model. For now, it detects potholes with bounding boxes, achieving a solid mAP of 65.5%, with 67.8% precision and 62.8% recall.👍👍👍 This is a work in progress, and I'm keen to circle back and perfect the edge detection. But to make this tool more accessible, I'm also building a simple web interface for it. It will support three ways to detect potholes: by uploading a picture, a video, or—the most awaited feature—using a live feed for real-time detection. I'll be cleaning up the entire project, including the web app, and will post it for reference shortly after it's complete. Stay tuned!!! ☺️☺️ #ComputerVision #AI #Python #YOLO #DeepLearning #ObjectDetection #RoadSafety #Infrastructure #Roboflow #GoogleColab #WebApp #RealTime #StudentProject

🚀 MagicEraser: Freeze Reality — My New OpenCV Project 🎥✨ Ever wished you could freeze time in a live video? ❄️ I recently built a fun and interactive “MagicEraser” project using Python + OpenCV, where you can pause the real world and selectively reveal or freeze parts of your webcam feed — all in real time! This project was came into thought by the post of Riyanshi Verma 🔧 Key Features: 🖱️ Brush over areas to freeze or reveal motion 🎞️ Toggle between "Freeze" and "Reveal" modes 🎨 Adjustable brush size for precision ♻️ Real-time video manipulation using OpenCV 💡 How it works: The program captures your webcam feed, freezes a selected frame, and uses a mask to blend live and static regions dynamically. It’s like painting time itself! 💻 Tech Stack: Python OpenCV NumPy This project helped me explore real-time image masking, bitwise operations, and frame blending — powerful concepts for computer vision and AR applications. 🌟 It’s a simple yet mind-bending demo that shows how creative coding can make technology feel magical. Would love to hear your thoughts! 👇 #Python #OpenCV #ComputerVision #AI #MachineLearning #Innovation #ProjectShowcase #Coding

Today, we're joined by Jacob Buckman, co-founder and CEO of Manifest AI to discuss achieving long context in transformers. We discuss the bottlenecks of scaling context length and recent techniques to overcome them, including windowed attention, grouped query attention, and latent space attention. We explore the idea of weight-state balance and the weight-state FLOP ratio as a way of reasoning about the optimality of compute architectures, and we dig into the Power Retention architecture, which blends the parallelization of attention with the linear scaling of recurrence and promises speedups of >10x during training and >100x during inference. We review Manifest AI’s recent open source projects as well: Vidrial—a custom CUDA framework for building highly optimized GPU kernels in Python, and PowerCoder—a 3B-parameter coding model fine-tuned from StarCoder to use power retention. Our chat also covers the use of metrics like in-context learning curves and negative log likelihood to measure context utility, the implications of scaling laws, and the future of long context lengths in AI applications. 🎧 / 🎥 Listen or watch the full episode on our page: https://twimlai.com/go/750.

🚀 Just built something exciting: Double Blink to Scroll using Python, OpenCV, and MediaPipe! 👀✨ Instead of touching your screen or keyboard, you can now scroll with a simple double blink. This opens doors for hands-free navigation, accessibility tech, AR/VR integration, and futuristic user interfaces. This is just the beginning — imagine what’s possible when AI meets human-computer interaction. ➡️ Accessibility for differently-abled individuals ➡️ Seamless browsing in AR/VR ➡️ Productivity hacks for the future I’d love to hear your thoughts and ideas on where this could be applied! 🔗 Built with: #Python #OpenCV #MediaPipe #AI #Innovation 💡 Dreaming big: Building technology today that feels like science fiction tomorrow. #ArtificialIntelligence #Accessibility #ComputerVision #FutureTech #MachineLearning #TechForGood #doubleblinktoscroll Meta | Instagram | Meta for Developers