This JavaScript AI Guesses Your Age & Emotion (Face-API.js)

AI Face Recognition in JavaScript: Testing Face-API.js & Fine-Tuning a Model

Explore our real-time facial expression and age recognition project built with Face-API.js and TensorFlow.js. In this video, we test a web application that runs entirely in your browser, analyzing your webcam feed to detect emotions like happy, angry, and surprised, while also estimating your age. But how accurate is it?

We put the model to the test and immediately notice some interesting quirks, especially with the age estimation which fluctuates wildly. We use an AI assistant (Gemini) to analyze the application's performance and architecture in real-time, discussing key questions: How can we improve the model? Where are the weights stored for a client-side application? Can we fine-tune the model directly in the browser to correct its mistakes?

We examine the proof-of-concept "fine-tuning exploration" feature, designed to let users train a custom mini-classifier. However, we run into some unexpected roadblocks. We discover that the fine-tuning is only for facial expressions, not age, and more importantly, the core functionality for capturing samples and training the model isn't even implemented yet!

Join us as we troubleshoot the application, discuss the challenges of client-side AI, and create a detailed developer brief outlining the necessary improvements. This includes implementing the fine-tuning buttons, adding functionality to save the trained model using browser storage like IndexedDB, and exploring ways to improve age detection.

All our projects, code, and prompts are available on BioniChaos.com. Go check it out and provide your feedback!

#FaceAPIjs #TensorFlowJS #AI #JavaScript #WebDevelopment #MachineLearning #FacialRecognition #ComputerVision #BiomedicalDataScience #AIinBrowser #CodingProject #BioniChaos #DevVlog

00:00 - Live Demo: Real-Time Face Expression & Age Recognition
00:23 - Using an AI Assistant to Analyze the Web App
01:13 - Analyzing the Model's Inconsistent Performance
02:10 - Can We Improve the Face-API.js Model?
02:57 - Exploring the "Bonus Fine-Tuning" Feature
03:40 - How Canvas Size Affects AI Detection Accuracy
04:45 - The Concept: Fine-Tuning an AI Model in the Browser
05:53 - Web App Architecture & Client-Side Model Storage
06:46 - Where are the Pre-Trained Model Weights?
08:53 - Discovery: Fine-Tuning is for Expressions, Not Age
09:37 - Creating a Developer Brief to Improve the Application
11:06 - Reviewing the Model's Strengths & Weaknesses
13:46 - Attempting to Use the Fine-Tuning Functionality
14:30 - Problem: The "Capture Sample" Button Doesn't Work
15:50 - AI Confirms: Fine-Tuning Buttons Are Not Implemented
16:29 - Summary and Next Steps for the Project
16:39 - Call to Action: Visit BioniChaos.com

Check out the tools we develop at https://bionichaos.com
Support BioniChaos on Patreon: https://patreon.com/bionichaos
Become a channel member to get exclusive perks: https://www.youtube.com/channel/UCCNBUsRwWl-J4di-GIZG16A/join

Видео This JavaScript AI Guesses Your Age & Emotion (Face-API.js) канала BioniChaos