Our Services

    Clients might draw on our expertise in machine learning, computer vision and language modeling for their challenges to build smart products or provide intelligent services. At Deep AI we focus on building high-quality software toolboxes and applications in the following areas:

      Smart Monitoring

      Our software allows for fine-grained identification of brands and products in any videostream enabling customers to track and monitor brand citedness in their ad campaigns. Thus for example sponsoring events can be automatically analyzed with respect to brand or product visibility and then be included in any program on evaluation of ad spending efficiency.

      Intelligent Surveillance

      We provide software tools to identify persons, animals, general objects and even complex actions on surveillance videos. With our software customers can avoid sifting through hours of irrelevant video material and directly get all relevant information from the surveillance stream to their dashboard or app.

      Visual Analysis & Recommendation

      Based on visual similarity of products our software can automatically select and recommend products for customers of your online store. Even beyond visual similarity of products it might be used to recommend products based on other visual cues taken from mobile cameras or web-cams. The software can be conveniently integrated into any existing web or mobile infrastructure.

    Our Approach

    At Deep AI we employ state-of-the art methods in deep learning such as very deep convolutional neural networks or advanved sequence labeling techniques to challenges in machine learning and vision / language modeling. Often we build on recent code from open source releases on github and adjust them to our tasks. For data analysis and prototyping we typically use Berkeley's Caffe, Google's Tensorflow or dmlc-mxnet. Our network models have often been trained on hundreds of Gigabytes or even Terabytes of data on our proprietary Maxwell and Pascal GPUs. We typically expose the resulting network models as RESTful APIs which clients can then integrate into their environment. For inference we take advantage of NVIDIA's GPU Inference Engine, for deployment we make heavy use of containerization with Docker (nvidia-docker).

    Our Vision

    Machines are increasingly capable to interpret our environment in a human-like way. This opens up limitless opportunities to advance existing products or build entirely new ones. At Deep AI we aim at facilitating that 'intelligentification' by making software tools available to any business whose products could benefit from smartly sensing its environment.

    At Your Service

    Building on our in-depth understanding of the many possibilities and a few limitations of current artificial intelligence we advise clients how their products and services could benefit from the infusion of artificial intelligence. To make our advise more tangible we often setup a prototype that is solving a partial client requirement. In a potential subsequent implementation phase we build a fullscale customized solution. This can be a tailor-made software module for in-house consumption or an exposed API in the cloud.

      How can your industry, your product or your service benefit from intelligently sensing the environment?

      Latest news on deep learning & artificial intelligence