📰 Medium · Python

A fixed stride straddles every cut. Scene cuts alone leave long shots unsampled. The fix is to let one decide structure and the other… Continue reading on Mediu

If you’ve ever wondered how Instagram applies filters in real time, how your phone unlocks with your face, or how a self driving car… Continue reading on Medium

How I replaced a repetitive laboratory workflow with an open-source robot, a computer vision pipeline, and a validated deep learning model Continue reading on T

A YOLOv8 + ByteTrack pipeline for real-time pedestrian counting, and the two bugs that almost made its numbers meaningless. Continue reading on Medium »

Counting reps sounds trivial until you try to make it work on a real phone video. This post walks through how I built a jumping jack… Continue reading on Medium

If you know me, you know I love VPython — the module that brings 3D graphics and other stuff into python. In short, this simple coding… Continue reading on Medi

As someone new to computer vision, I wanted to build a project that would expose me to the complete deep learning workflow — from… Continue reading on Medium »

Why LivePortrait glues an animated face back onto still shoulders with three small MLPs that predict keypoint deltas, and what that… Continue reading on Medium

Explore how Vision Transformers are redefining the future of computer vision. Continue reading on Medium »

One curated path through 3D reconstruction, from classical photogrammetry to AI depth to Gaussian Splatting, so a folder of ordinary… Continue reading on Medium

“YOLO finds the object, while OCR reads the text on it. Powerful on their own, these two technologies become the foundation of intelligent… Continue reading on

In most computer vision tutorials, everything starts with a clean, pre-annotated dataset downloaded from the internet. But real-life… Continue reading on Medium

Practical architectures for video recognition in PyTorch and Hugging Face without the compute cost of 3D CNNs. Continue reading on Medium »

Imagine you’re building an AI system that detects cars in a parking lot. It successfully draws bounding boxes around every vehicle, but it… Continue reading on

Part 1 of this series built an image preprocessing kernel where every output pixel maps to a predictable source location on a regular grid… Continue reading on

You have a Hilook or Hikvision camera and want to programmatically pull images from it? This short guide would help you do exactly that. Continue reading on Med

Why erasing a sign from a video is a masking problem first, an inpainting problem second, and why the detector you reach for matters more… Continue reading on M

Open any self-driving demo and somewhere underneath there is a neural network the size of a small country, trained on more video than you… Continue reading on M

You have a Tapo camera and want to programmatically pull images from it? This short guide would help you do exactly that. Continue reading on Medium »

What hand tracking and a little geometry can do that a segmentation model can’t Continue reading on Medium »

“Detecting an object is not always enough. Sometimes, the object needs to be clearer, cleaner, and easier to read before it can be… Continue reading on Medium »

“Training a YOLO model is only the first step. The real challenge is correctly interpreting the results after training to understand how… Continue reading on Me

Passive CCTV is failing residential communities. Here is how edge computer vision and polygon zone-enforcement turn standard cameras into… Continue reading on M

We expect our cameras to protect us, yet they do nothing but record while we play the security guard. Outsourcing this to AI makes sense… Continue reading on Me