Open source Computer Vision toolkit
Open source and easy to use. Install it in your own servers and integrate it with your product right away.
Easily train neural networks to detect and classify objects with custom data. Use state of the art models such as Faster R-CNN.
Built on top of Google’s Deep Learning framework and DeepMind’s graph library.
Train your model by just typing lumi train. Do it locally or using Luminoth's built-in Google Cloud Platform support to train in the cloud.
Once training is done, you can use our Tensorboard integration to visualize progress and intermediate results. Also, evaluate using different data splits.
The ability to visualize results is always important and more so in the Computer Vision field. After training a model you can get an easy to understand summary and image visualizations to spot results, using either our UI where you can adjust the probability threshold on the fly, or the command line interface.
Since the release of the library, the Luminoth core team has been invited to talk about Deep Learning for Object Detection in several events. The talks outlined the challenges, inner workings and learnings from building a toolkit.
Find next a talk titled "Building an Object Detection toolkit with TensorFlow: From academic papers to open source implementation". It was hosted by Alan Descoins (Tryolabs CTO) and Javier Rey (Tryolabs Lead Research Engineer) and given at the ODSC London.
