The Media’s AI

The Platform that simplifies AI

You Are in good company

Telekom

Visual Data Mining

DeepVA offers multiple visual recognition capabilities to extract any kind of information from images, videos and live streams. Our pre-trained AI models were developed in close cooperation with the media industry and strive for accuracy without compromising speed. We keep an eye on the changing environment and constantly provide flexible AI solutions for upcoming challenges in the industry.

Multi-Language Support

Frame Accurate Results

Frequent Model Updates

  • Faces
  • Face Attributes
  • Diversity
  • Visual Concepts
  • Landmarks
  • Name
  • Text
  • Shot Boundary
  • Teaserframe
  • Created by potrace 1.16, written by Peter Selinger 2001-2019 QR-Codes
  • Aspect Ratio
DeepVA face recognition

Content producers and TV broadcasters nowadays find it even more important to produce high-quality news articles and video content efficiently. The importance of recognizing people correctly is therefore essential for both editorial and production staff. DeepVA’s Face Recognition is designed to recognize all important public figures, such as people from politics, sports, arts, entertainment and business – for Europe, North America and the rest of the world.

Over 21.000
personalities

Constantly
improving library

99,36% accuracy
in identification

DeepVA Face Attributes Detection

Small things matter! Pay attention to the little details and analyze emotions like “happy”, “sad”, “angry” or “confused”, filter out odd images due to unintentional attributes like “closed eyes” or “open mouth” and detect several other characteristics as well as demographic information and much more! Always find images and videos for the right mood! It’s the perfect addition to Face Recognition!

Analyze emotions,
age & gender

Demographic analysis included

Facial attributes like “sunglasses” etc.

deepva diversity analysis

DeepVA’s Diversity Analysis was developed to provide a solution for more diversity in the media and helps media companies to take this into account already when creating content. This module allows for diversity monitoring in all your contents and lets you gain important insights via reports.

It is always about equal opportunities and social cohesion, especially in the face of increasing social changes. 

Gender, Age
& Ethnicity

Meet today’s expectations

Create diverse
media landscape

Diversity
reportings

Deepva visual concepts

Recognizing visual concepts in videos and images is important to help media companies effectively repurpose the content they produce. 

DeepVA’s visual concept recognition identifies objects and scenes to specifically help you automatically summarize, categorize and archive media content. With this module, you get a pretty good understanding of what’s going on in your images and videos with their corresponding timestamps.
Over 1,500 media-relevant classes

simple, frame-accurate labeling

Perfectly solution for your MAM/DAM

DeepVA Landmark Recognition

Being able to recognize and locate landmarks is essential in media broadcasting. DeepVA’s landmark recognition can help you identify the most important landmarks, architectural structures and natural monuments throughout Europe and North America. In addition, it provides you with geodata as part of the results. This module facilitates archiving and retrieval of landmarks.

Over 30.000 landmarks

Find the location
on the map

Easy archiving
and retrieval

Deepva Face recognition

DeepVA name recognition automatically detects and extracts names, identifies their role, and the geographic location of on-screen name inserts.

Name recognition can be a good extension to face recognition in order to find out who is actually speaking.

Identify
speakers

multi-language
support

Complements speech to text

deepva brand and logo identification

Monitoring, monetization, sponsorship measurement, counterfeit detection and brand protection are just a small selection of essential activities in today’s complex media landscape. Brand & Logo Detection is the ideal basis for all these activities and is extremely important for maintaining a clear overview at all times.

Perfect for
brand monitoring

Over 10.000
Logos

Maximize
awareness

Deepva Text recognition

Text or text overlays contain relevant information and occur in videos and images in large quantities. Therefore, it is inevitable that textual information must be recognized in real images. 

DeepVA’s Text Recognition recognizes and parses any information that is shown on the screen in images and videos. This can be an overlay, text on billboards, signs, documents or any other surface, no matter which font or whether it is handwritten. 

Extract text
and numbers

Perfect for digitalization

Multi-language Support

Deepva shot boundary detection

Feature films, series, documentaries, sporting events and news are edited together from many different shots. DeepVA’s Shot Boundary Detection helps you see the complexity of a shot list by detecting every slightest change.This makes it possible to skip intros, credits, title sequences and recaps, or find scenes and sequences that have attracted the most interest.

Video
segmentation

Make editing more comfortable

Improve consumer experience

Deepva Teaserframe detection

With the increasing popularity of streaming portals and social media platforms, the importance of videos is growing. The perfect representation of the content via a thumbnail is therefore essential. With DeepVA’s teaser frame extraction, you don’t have to watch an entire video just to find the most interesting snippet. Based on the key elements of the video, we select the most promising snippets for you. It doesn’t matter in which aspect ratio you need this preview frame, because we crop it so no important content gets lost.

Several thumbnails to choose from

Increase user engagement

Smart
cropping

DeepVa Qr code detection

It can be tricky to decode QR Codes in media content, especially if it is only displayed for a short period of time. We decode every QR Code that is displayed and also extract barcodes (EAN13) and the corresponding product names. Use QR codes in your content production to consistently match your media and keep your workflow consistent and meaningful.

Decode any
QR code

Allocate visual content

EAN code
extraction

deepva aspect ratio identification

With the increase of different devices and social media platforms, it has become more and more challenging for media companies to provide the correct format to their audience.

DeepVA’s Aspect Ration Detection detects all active aspect ratios in videos and images, even if the footage contains letterboxes or pillar boxes, and crops the content smartly and accurately to meet your needs.

Detect video
& image formats

Smart
cropping

Define the
best format

Custom AI models

DeepVA makes AI training predictable and easy to budget.

Trained in a few Seconds

Ad Hoc 
Training

3 Clicks
Training

Only 1 Sample Image Needed

  • Custom Face Model
  • Custom Landmark Model
  • Face Indexing
deepva custom face model

Pre-trained face models quickly reach their limits and, in most cases, do not reflect the needs of the media industry. With the Custom Face Models, DeepVA guarantees the possibility to train individuals and especially region-specific personalities. With Few-Shot Learning, media companies are now able to build their own face models in just a few seconds.

Few Shot
Learning

build your own
Face Recognition

Combine it with our
pre-trained models

deepva custom landmarks creation

Recognizing buildings creates a huge editorial benefit for media companies in their press coverage. With DeepVAs Custom Landmark Model there is now the possibility to easily extend our Landmark Recognition. You are now able to recognize your most relevant landmarks with just a few clicks.

Few Shot
Learning

your own
Landmark Models

Combine it with our
Pre-trained Models

deepva face indexing

Face Indexing is another great way to recognize faces. This state-of-the-art method of face fingerprinting allows media companies to analyze media data followed by assignment of unique IDs. Thus, unknown persons can be described retrospectively across all your media data without the need for further analysis.

Similarity
& Reverse Search

Detection
of Duplicates

Ad hoc
training

Dataset

Management

Dataset Management

Datasets are essential for training and customizing Al models. However, the amount of data can quickly get overwhelming and time-consuming. DeepVA ensures that handling your data is transparent and intuitive.

Auto Quality Checking

Auto Data Collection

Smooth Data Handling

Easy Dataset Export

Helps Building Diverse Datasets

  • Face Dataset Creation
  • Face Dataset Evaluation
  • Dataset Management
deepva face dataset creation

Face Dataset Creation is the perfect tool if you don’t have time to collect faces from people yourself. It searches for important people in videos, crops their faces, checks their quality and stores the data in labeled datasets. 

Everything is ready to start training your face recognition model.

smart face cropping
from videos

name class
mapping

extracts most important persons

deepva face dataset evaluation

High quality data is essential for AI training. It is very time consuming to build your own datasets and clean them afterwards.

Face Dataset Evaluation supports the user in building high quality face datasets and evaluates the data to check for variance, diversity and errors.

Detects Face Masks, blurry faces etc.

Outlier Detection
& Clustering

Data
Improvements

Diversity
checking

deepva dataset management

There are many challenges in building, managing and maintaining AI datasets. DeepVA therefore provides features that make it easy for you to efficiently manage very large datasets while providing all the functionality to maintain data privacy.

Efficient handling
of large data sets

Dataset export options

Advanced privacy features

 

DeepVA offers excellent solutions for AI-based content recognition for the media industry. The modules such as face and object recognition, as well as emotion and demographics recognition are essential for industrial application. Compared to the other AI software providers, DeepVA scores with a wide range of services. Another big advantage is that this software runs on a cloud server in the EU, which means there will be no problem meeting the increasingly strict European Privacy Policy regulations as well as the upcoming Digital Markets Act. With DeepVA as a partner, we believe we can perfectly meet all the necessary data safety requirements.

Ralf Hülsmann

Public Cloud Managed Services – Strategy & CTO Office

Get Started Right Away

Get Started

Right Away

On-premise

If you have strict data protection rules or can’t transfer your data to the cloud, simply deploy DeepVA on-premise and virtualize it with Docker, Kubernetes or VM-Ware. 

Cloud

DeepVA can run on every Cloud, that is supporting Docker or Kubernetes like AWS, Google Cloud, Open Telekom Cloud or Microsoft Azure.

Self-hosted or managed

Choose whether to manage DeepVA yourself, or have us help you operate the AI platform.

Updates and fixes

DeepVA is improving on a regular basis. On the one hand, our AI algorithms are getting better, on the other hand, we are still adding concepts or people to our models.

Partner Integration

The media landscape is complex and each media company has specific needs and requirements. That’s why we work closely with suppliers and representatives of the media industry to provide media companies the best possible support. Become a DeepVA partner and benefit from an innovative and valuable cooperation.

Our experience with generic AI solutions is that the desired entities are often not recognized in their own content. Creating training data and managing datasets is complex and time-consuming. DeepVA's integration with our media service has shown us that customizable AI within the already familiar MAM-interface can be intuitive and straightforward.

Ralf Jansen

Software Architect at Vidispine
– An Arvato Systems Brand

Ralf Jansen Vidispine

DeepVA's approach to integrating AI into existing MAM-systems really excites me. This way we can make artificial intelligence accessible to smaller media companies. Behind DeepVA is a small team that is results-oriented and professional, while remaining flexible and customer-focused. We enjoy working with this small innovative start-up from southern Germany.

Marc Jonas

CEO, Nachtblau GmbH

Our experience shows that working with DeepVA always leads to excellent results. This was made possible not only by DeepVA's expertise in AI and experience in the media industry, but also by the straightforward and agile collaboration. The comparatively unprecedented licensing model, which is ideally suited to the needs of media and entertainment companies, is worth highlighting.

Matthias Naumer

CEO, Munich Media Intelligence GmbH

Matthias Naumer CEO MMI

Data Security & Privacy

Data Security

& Privacy

Deepva cyber security datenschutz

Your Data Belongs to you

The sheer number of images and videos makes it difficult to ensure data security and privacy. Whether you use DeepVA on-premise or in the cloud, we guarantee that your data always stays in your control. We promise full transparency and follow the latest compliance guidelines. All our cloud servers are located in Europe and everything we do is 100% GDPR-compliant.

  • How does DeepVA help you stay GDRP compliant?

    With our intelligent cloud management, DeepVA offers you the possibility to decide where your data should be processed. Basically, our servers run in German and other European data centers. Furthermore, you have the option to use our technology on Prem at any time. Either way, you are in full control of your data.

  • Human in the Loop

    DeepVA offers you many possibilities to automate your media workflows. However, data protection plays an essential role, so we provide you with tools that help you maintain full control over the process. This way you can ensure that no data is unintentionally collected that can be used for any kind of analysis.

  • Anonymization of personal data

    Our AI is designed to store the results of personal data from unknown people only temporarily and as pseudonyms. You have the option to automatically delete irrelevant or rarely occurring persons from your dataset.

  • You have sovereignty over your datasets

    All your uploaded data used for training or easier management will be strictly withheld from DeepVA or any third-party service. This applies to any of our provided licenses.

We Are supported by