Efficient^AI

Efficient^AI is the groundbreaking data-efficient artificial intelligence technology for industry-grade applications.

DATA EFFICIENT &
FLEXIBLE

Five to ten image samples are sufficient to learn and adapt visual classification tasks. Add new classes whenever you need them.

“10 not 100 Training Samples”

INTUITIVE &
IMMEDIATELY APPLICABLE

No need for expert knowledge, time-consuming calibration, or (re-)training process. Adjust promptly to your demands.

“1 s to Teach — 1 s to Classify”

EXPLOIT UNTAPPED
INDUSTRIAL POTENTIAL

Enabler for novel applications, delivered in industry-grade systems or as a standalone software – even applicable from remote.

“Infinite
Potential”

A.I. meets modern industrial demands

Current AI-based visual classification technologies typically excel the performance of classic algorithms, but require massive amounts of training data and an extensive training process before being able to put into the field. This is an insufficient approach for industrial applications that rely on repeatedly changing (product) portfolios or lack of pre-existing image datasets.

Efficient^AI solves this data-hunger and inflexibility by providing a novel machine learning approach based on a groundbreaking few-shot-learning technology. Moreover, it could be instantly used and ensures cost and time saving for maximum performance of vision applications in the industrial context.

Application domains
EfficientAI is camera- and hardware-agnostic and serves different applications domains of visual classification tasks. 

Quality Control & Inspection

Inline or end-of-line inspection and quality control systems secure the highest standards of manufacturing and assembly and help to reduce waste and overall production failures.

The complexity of modern production makes it difficult to ensure the economical use of modern AI, and even failures should be detected that are not known before or occur very rarely.

Efficient^AI copes reliably and robust with a wide variety of visual quality features and is adaptable within minutes. In addition, it is extendable toward anomaly detection.

Inventory & Flow Assessment

Monitoring and assessing the material flow in modem production and logistics environments is a core need for process optimizations and just-in-time deliveries.

However, changing product portfolios and dynamic environments are still a major hurdle for modern AI approaches as those demand huge datasets and extensive training.

Efficient^AI is the strong basis for modern visual check-in/out and material flow assessment systems that efficiently adapt to changing product portfolios and product variations.

Sorting, Packing & Picking

Flexible (robot-based) handling for sorting, packing, and order picking requires the reliable identification of objects and parts before pose estimation and grasping.

Especially if the variety of target objects is wide or changing over time, data acquisition for a common AI approach becomes expansive and time-consuming.

Instead, Efficient^AI is trained with just 10 sample images per part and successfully proofed full performance in robot-based sorting cases with >350 different objects.

efficientai ui pcb — EfficientAI’s intuitive user interface guides the user in a structured fashion through data acquisition and the instant classification results.

Efficiency as a Service

Award-winning Efficient^AI is distributed as a

software as a service module,
standalone system including hardware or
customized full service system solution including integration.

We offer continuous support and work with hardware partners to provide suitable equipment.

Deep Tech Award Winner
More details at our blog

inVISION Top Innovation
More details at our blog

Efficient^AI and its simplistic process

“Learn and Measure Differences — Instead of Expensively Learning Overall Concepts”

Solve data hunger: train on non-task-specific data – apply on a little task-specific data
Solve compute hunger: train once – apply on arbitrary new tasks

Talks

Computer Vision Applications: Case Studies from Germany, Poland, USA

Organizer
Digital Business Association Berlin-Brandenburg (SIBB)

Details
10/2021 | 15 min online presentation | Computer vision, Digital showcase

YouTube-Feature

NVIDIA GTC 2018 Technology Presentation: Metric learning to reduce deep learning's Resource Hunger

Event
NVIDIA GTC conference

Details
10/2018 | 35 min technology introduction | Reduce data-hunger, reduce computing-resources, train once, use forever, embedded hardware, classification

Presentation (PDF) | YouTube-Feature