#tensorflowjs

Machine learning using TensorFlow for Absolute Beginners https://morioh.com/p/4dd4d14a2b7e

The AI Revolution in Web Development

At Prishusoft, we recently delivered a powerful AI-driven web application using TensorFlow.js, enabling real-time machine learning right in the browser, with no server-side dependency!

Real-time predictions

No downloads, no delays

Cross-platform performance

Fully private, your data never leaves the browser

Whether it's image recognition, predictive modeling, or interactive ML features, our team is ready to bring your ideas to life using TensorFlow.js.

If you have a project or idea involving AI, ML, or TensorFlow.js, let's talk! Visit: https://lnkd.in/gE9RhPAd

we’re ready to help you build smart, scalable web solutions.

Last week I discovered my garage door was left open all night because my kids were playing outside the day before and forgot to close it. That was a serious security issue that needs to be rectified. So I was on a mission to solve this problem and started the safer garage project.

First I wanted to install a wifi security camera. I mounted the camera facing the garage doors so if I was concerned I could always launch the app on my phone to check the door. This was a good solution but not foolproof since I can (and probably will) forget to check every day.

What I need was something to notify me when the garage door was left open.

So I looked into coding up a solution.

Fortunately, for me, my cheap Amazon camera is capable of RTSP which is a standard protocol and so I can easily pull video/images from it. Now I needed some mechanism to recognize when the door was open and when it was closed. I had an inkling I needed some employ machine learning to accomplish this so I started googling. I discovered OpenCV, and TensorFlow and stumbled on what looked like an easy-to-use web app called Teachable Machine which is built on top of TensorFlow.JS.  It's in JavaScript something I am familiar with.

Teachable Machine was great in that it's easy to use to get started. I started feeding it various screenshots of my garage door in various states (with the door open and closed and under different lighting conditions) and right away I got some very promising results.

So onward!

I downloaded the Model TM generated and tried to open it in node. Unfortunately, TM is entirely browser-based and so I was forced to do some hacking involving jsdom, and node-canvas. This was probably the ugliest part. Anyway, it's done. I started looking at pulling the images off my webcam and came upon FFmpeg which is an application that's been around forever and is pretty much the cURL of the video world. Spawn this app and get an image. Done and done! So now I have it hooked up together and it sends me a push notification when the garage door is left open.

Problem solved!

Build A Machine Learning Image Classifier with TensorFlowJs ☞ https://morioh.com/p/b468f80e2d0b

How to Categorize TensorFlow.js Images Made easy ☞ https://morioh.com/p/4891081d82c3?f=5c21fb01c16e2556b555ab32