A basic web application that identifies objects through your webcam. My goal is to familiarize myself with Tensorflow's Object Detection API.
This code loads the required libraries from Content Delivery Networks, so you need not install them on your local machine.
- git clone this repository:
git clone https://github.com/sarahdahippo/live-recognition.git cdto thelive-recognitiondirectory- run
python -m http.server - navigate to http://localhost:8000/ in your web browser
- allow webcam access when prompted
- there might be some delay time before the model starts detecting objects
Tensorflow.js is a library for developing Machine Learning models in JavaScript, and use it in Node.js or directly in our browsers. With this, I can utilize pre-existing models and retrain them with my own data.
COCO-SSD (Common Objects in Context-Single Shot Detector) is an object detection model that detects and localizes one or more objects in a picture. COCO encompasses all the data (images) that trained the model. Currently, there are 80 different classes that the model can detect.
Train my own custom object detector using TensorFlow Object Detection API, then convert it using tensorflowjs_converter.