DeepTeach is an MLDB.ai plugin that allows a user to teach a machine learning model what types of images he’s looking for through an iterative process. It's a great example of human augmentation, where machine learning is used to make humans more efficient.
The plugin uses the Inception-v3 model, a deep convolutional neural network, as its feature generator. It then uses the user's input to train a bagged boosted decision tree in order to learn the type of image the user is looking for. It's a combination of active learning, deep learning, transfer learning and similarity search.
Some links:
Try DeepTeach for free! Just create a free MLDB.ai account to launch an instance and run the Transfer Learning on Images with Tensorflow demo from within your MLDB instance.
One way is to the bottom of the Transfer Learning on Images with Tensorflow demo notebook from a running instance of MLDB.
Alternatively, from a notebook running on MLDB, run the following:
from pymldb import Connection
mldb = Connection()
mldb.put("/v1/plugins/deepteach", {
"type": "python",
"params": {
"address": "git:https://github.com/mldbai/deepteach"
}
})
You can then browse to https://<host:port>/v1/plugins/deepteach/routes/static/index.html
to access the UI.