Who is the WEAVER?

Adaptive Vision logo

The WEAVER is a new deep learning inference engine that comes with Adaptive Vision Studio 4.12. He is user friendly, cares greatly about performance and reliability. This newsletter will help you to get to know him better.

1- Performance

Deep learning networks are typically created with frameworks like TensorFlow, PyTorch or Caffe. These are great environments for research. However, to achieve the highest possible performance you will also need an inference engine dedicated to your hardware platform. Our experiments have shown that relatively mature and usable choices are: TensorRT (GPU), OpenVINO (CPU), MXNET (GPU), PlaidML (GPU) and ONNX Runtime (CPU). WEAVER is a new option. He works best in machine vision applications and provides state of the art performance on both GPU and CPU.

Below you can see how WEAVER‘s performance looks like on our most popular neural network, Feature Detection (supervised pixelwise labeling, Thread Inspection official example):

Feature Detection: Performance on GPU GTX 2060
Feature Detection: Performance on GPU GTX 1060

On nVidia RTX 2060, it gets even more interesting. TensorRT can go faster when numeric precision is reduced to 16 bits, while WEAVER still works with full 32-bit precision:

Feature Detection: Performance on GPU GTX 2060

On the Intel CPU platform we compare WEAVER against OpenVINO and ONNX:

Feature Detection: Performance on CPU
Feature Detection: Performance on CPU

DISCLAIMER: The WEAVER inference engine was specifically optimized for neural network we use, including Feature Detection. Performance for other neural network architectures may be different.

2. Reliability

The fast pace of development in the field of deep learning is not particularly helpful in developing real-world systems. Most of our customers require long-term support and stable availability of the components they use. WEAVER solves that. He is independent of the open-source frameworks, eliminates Python code and provides long-term support for commercial projects.

3. Ease of use

Most of the other inference engines require you to do the Python programming and tweak many things. WEAVER is different. He only does two things: (1) model optimization, (2) execution. All you need to deliver is your H5 network file (Keras output). Moreover, when you create your solution with Adaptive Vision Studio, WEAVER is already employed in three ready-to-use deep learning tools: Feature Detection, Point Location and Object Classification.

Model Optimization and execution

DISCLAIMER: TensorRT, OpenVINO, MXNET, PlaidML, ONNX, TensorFlow, PyTorch and Coffe are trade marks of third party companies.

Blog by: Adaptive Vision


EyeVision software from EVT now offers EyeMeasureEasy (EME), an intuitive image analyzing tool. The user can directly measure and analyze image data with only a few mouse clicks. The tool can be used for applications in microscopy. The features of EME can be extended, so that it can be integrated into a fully automated, industrial inspection process.

With EyeMeasureEasy the user can display live-images, and can save images and load them from memory. There are several different measurement options that can be carried out on those images, such as:

  • Angle including three points or four points
  • Distances between two points
  • Radius of circles including three points or n-points or automatic

The measured results can be shown via a text overlay, list boxes or charts. The measurements can be controlled by the user through “start” and “stop” buttons, or the measurements can be carried out in a free-running measurement.

The camera options can be set directly and the user interface is a design framework for a customized graphic user interface.



BGA Inspection with EyeScan AT 3D from EVT

BGA Inspection with EyeScan AT 3D from EVTThe EyeScan AT 3D sensor – together with EyeVision image processing software can inspect the correct position of pins and check for the quality and position of balls on a Ball-Grid-Array. The connectors of BGAs can be inspected with the 3D sensor for various characteristics up to an accuracy of a micrometer (µm).

Thanks to the Scheimpflug-projection, the laser line stays focused and distortion-free. Even in a high-speed application, the captured images have a high resolution, and no lens distortion correction is needed. Every single bump of the BGA is inspected for its size and shape up to a µm-accuracy.

The EyeScan AT utilizes precision laser line measurement (or laser triangulation). The camera views a laser line, which is projected onto the component (in this case: the BGA) and calculates the height information from the laser line profile.

Utilizing a precision blue laser and on board FPGA the EyeScan AT 3D is capable of over 20,000 profiles per second with very high resolution to capture even the finest details. Additionally, the High Dynamic Range algorithm with threefold exposure allows inspection of bright and dark objects in the same scan without loss of speed.

3D image information is processed by the powerful EyeVision software which is programmed though an intuitive drag and drop interface.

EyeScan AT 3D with Blue Laser Line

EyeVision software by EVT now also supports a 3D laser triangulation sensor with blue laser. The new EyeScan AT 3D scanner is especially suitable for scanning metal or metallic surfaces.

AT3D blue BrainLight ringToday’s noncontact 3D scanners include one of two laser types: blue laser or red laser. Most 3D scanners, including EyeScan 3D scanners from EVT, user red laser technology. The new EyeScan AT 3D Blue, however, utilizes optically superior blue laser technology.

The blue-violet Laser has some clear advantages which provide better scan results. Because it has a shorter wavelength than a red laser, a blue laser delivers improved scanning results with higher resolution and up to a 50-percent reduction in data noise. The shorter wavelength creates a smaller laser speckle pattern, generating a sharper depth image. The blue laser is more effective when scanning metal, because a blue laser is narrower and finer than a red laser line.
Thanks to 20,000 profiles per second and the new blue laser, even complex component parts and surfaces can be captured with high accuracy. The extremely fast scans are possible due to the short exposure time and the preprocessing of data with a FPGA inside the scanner. The high dynamic range algorithm with triple exposure makes it possible for dark and bright objects to be inspected in one scan, without a decrease in speed.

EyeCheck Smart Camera – the Eye for the Robot

Throughout the world robots are used for applications such as packaging, soldering, adhesive bonding, coating, etc. The robots are usually equipped with smart cameras, such as the EyeCheck 1000 series from EVT. Due to the recent advances in image processing, application areas have increased.

Roboter_mit_EyeCheckEyeCheck Smart camera series from EVT offers a combination of a camera and Eye Vision evaluation and processing software. The camera housing contains the lens (the user can chose between 6 mm, 12 mm and 25 mm lenses) and the illumination (available in different colors). The EyeCheck 1000 is available in various resolutions, from 782 x 485 pixel up to 1280 x 1024 pixel, and contains the full command set of EyeVision software.

The EyeCheck systems can be used for robot positioning and object detection, custom-made solutions for position correction, and can be mounted onto articulated robots. The moving or stationary objects are illuminated with the camera-based LEDs. Thanks to the robust evaluation algorithms, EyeVision software delivers reliable results even in difficult environments.

EyeVision software also has commands for code reading, color and surface inspection as well as commands for measuring objects. The system is configured and controlled by the user via the graphical user interface. The drag-and-drop programming allows an easy adjustment to different objects. EyeVision software can be connected to an external computer or the control system of a plant via Ethernet or PROFINET. Applications such as position control can be used in combination with inspection, measurement and identification tasks.

EyeVision 3 with New PlugIn System

PlugIn_puzzle-KopieThe EyeVision 3 machine vision software is more flexible than ever. With the new plugin concept, third party software such as Halcon, Open CV or the Euresys library can be integrated into the EyeVision software library.

Using the new SDK for plugin development, it is easy to integrate evaluation routines into EyeVision. EyeVision supports x86 Windows, Linux, as well as ARM based Linux systems.

With the GUI designer and only a few mouse clicks, the user can open their own routines or the library routines of a third party maker without special programming skills.

Also the standard communication components allows easy integration into SCADA and PLC systems. This means, for example, that the OPC standard can use EyeVision software as an OPC server. With that the accumulated data (measurement data, position, etc.) can be sent to a client.

Additionally, the new Webshop from EVT makes it possible to offer plugins directly to other EyeVision software users.

Crown Cap Inspection with EyeVision Image Processing Software

Crown Cap Inspection with EyeVision Image Processing Software

EyeVision software contains commands for various measurement technology applications, for pattern matching, for code reading, for object counting as well as error detection such as scratches, holes, cracks, contaminations, printing or color errors, etc., which makes it ideally suited for use in the bottling industry.  

Crown Cap Inspection with EyeVision Image Processing Software

EyeVision software can inspect the fill level, detect if bottles are without caps, or check the correct position of the label. It also offers the opportunity to inspect using 3D, for example, to check that the cap is installed properly.  

EyeVision software supports various 3D sensors such as the EyeScan AT 3D and contains commands for evaluating point clouds and grayscale images. The EyeScan AT 3D has a profile resolution of up to 4096 points/profiles (4k Ultra-HD) and integrated 3D evaluation algorithms with up to 25 kHz and a robust enclosure with IP67 and M12 connectors.

FSI Advanced Research will be exhibiting at the 2016 SC Manufacturing Conference and Expo

FSI Advanced Research will be exhibiting at the 2016 South Carolina Manufacturing Conference and Expo

FSI Advanced Research will be exhibiting at the 2016 South Carolina Manufacturing Conference and Expo, April 20-21, at the TD Convention Center in Greenville, SC. Join manufacturing professionals from across the Southeastern U.S. to discuss the latest manufacturing trends and topics, and immerse yourself in manufacturing culture. Also stop by booth #122 to learn the latest in machine vision technologies from FSI Advanced Research, including 3D and Deep Learning.