Peeking into the Neural Network Black Box

If you have ever come in touch with neural networks, you are probably familiar with the black box problem [1, 2]. Compared to many other algorithms from the glass box…

Tensorflow for cropping and rendering

In tensorflow, there are many usable features for tasks besides just for learning. One worth mentioning is the function for rendering framed rects for given boundingboxes that can be used…
Does Adding One Neuron Help Real World Networks?

Does Adding One Neuron Help Real World Networks?

Our researcher, Tonda Hoskovec, has long been thinking about the behavior of neural network training in the case of non-convex tasks with many local minima. This makes training difficult or…
How to automatically read data from invoices using Rossum API: an example with Google Drive

How to automatically read data from invoices using Rossum API: an example with Google Drive

Our research programmer, Roman Sushkov, recently posted a neat example of how to use Rossum’s Data Extraction API in practice. We think it is a great introduction to practical usage of…
Why did we start Rossum?

Why did we start Rossum?

Hear from Rossum’s Co-founder and Chief AI Architect Petr Baudis about our motivation behind the Rossum technology.
The Future of Data Capture Systems (2/2): The Rossum Approach

The Future of Data Capture Systems (2/2): The Rossum Approach

This is the second part of our special founders blog post on data capture technology and how Rossum represents a radically different approach to the whole problem of information extraction…
The Future of Data Capture Systems (1/2): Imitating Human Behavior

The Future of Data Capture Systems (1/2): Imitating Human Behavior

Data capture for invoices ought to have been solved a long time ago! That’s what most people think, especially if they’ve never tried to actually do it. That’s what we…
Towards Efficient Multi-GPU Training in Keras with TensorFlow

Towards Efficient Multi-GPU Training in Keras with TensorFlow

At Rossum, we are training large neural network models daily on powerful GPU servers and making this process quick and efficient is becoming a priority for us. We expected to…
Data Hackathon at our Prague Lab

Data Hackathon at our Prague Lab

At Rossum, we just talked about how critical data is for machine learning and for us specifically. At the same time, we’d like to think we are doing some pretty…

1 2