This repo is my stash of code examples using the R package Rvision. Rvision is a really great package out of the Swarm Lab at the New Jersey Institute of Technology. This package ports aspects of the massive OpenCV Library for machine vision into R; a new and needed addition! I started experimenting with this package soon after its release and found it very fun to work with. This repo contains my code and examples of working with the Rvision package in conjunction with Tensorflow and Keras machine learning libraries.
The basis for much of this code is a series of tweets and discussion with the Swarm Lab PI and primary package author Simon Garnier. I am not affiliated with the Swarm Lab or the Rvision package. Also note, the examples here have a number of dependencies that require some attention. Rvision itself requires the Rtools tool-chain and CMake. The Tensorflow and keras installs range from pain-free to a total pain in the butt. See the package links above for installation details.
There is still a ton of work to be done here and these examples are simplistic, but enjoyable to put together.
- Video_balloon_blob.R - single object tracking example
- MNIST_CNN_video_predict.r - use CNN trained on MNIST to classify numbers in video
- LIVE_webcam_CNN_single_blob_prediction.r - NOT FINISHED - object tracking and CNN prediction with ResNet50
- LIVE_webcam_CNN_prediction.r - ResNet50 CNN classification of video frames
- LIVE_webcam_blob_detection.r - object tracking from webcam
- KERAS_train_MNIST_CNN.r - script to train MNIST CNN with Keras (from Keras examples)
- LIVE_webcam_MNIST_CNN_predict.r - webcam stream with MNIST trained model - experimental...