Image Recognition with Deep Learning From prototype to production
summary
A clothing recognizer trains a convolutional network on about 300,000 labeled images using Python with Keras (TensorFlow backend) and one GPU, then truncates the classifier to extract ~1,000-d feature vectors and serves similarity by k‑nearest neighbors (scikit‑learn). A store demo wires a Kinect v2 to a Node.js worker, streams body/hand JSON into a React + Three.js frontend wrapped in Electron, uploads photos to S3, and accepts circular “hold to click” gestures; the system is deployed in stores and runs after engineering fixes.
AI-written summary · about
chapters (21)
- 0:00Intro
- 1:18What we've built
- 2:25Deep Learning
- 4:35Training your kid (or neural net)
- 6:17Convolutional Neural Networks
- 7:26How we trained our model
- 8:51Creating the model
- 10:24Training the model
- 12:22Categorizing new images
- 13:16Categorizing images
- 15:17From categorization to recommendation
- 17:50Lessons Learned
- 31:16The idea...
- 32:02Steps to take...
- 32:54Putting it all together...
- 34:31How was this made?
- 35:22How was this made: data
- 37:03Clothing Analyzer - Difficulties
- 38:07Clothing Analyzer - First test...
- 38:59Clothing analyzer - Version 1
- 40:01Clothing analyzer -Solutions (V2)
related talks
24mImage Recognition for safety on the factory floor
26mWhere can I park? Deep Learning is here to help you!
42mDeep Learning for Computer Vision
01h 04mBuilding and Training your own Custom Image Recognition AI
42mProduct Recognition in Retail/E-Commerce with Azure AI
21mRecognizing Human features using Deep Networks
15mMy first steps on Deep Learning with Keras
40mHow to Implement a Production Ready Deep Learning Model in 12 Weeks