AI-Solution

Advanced technologies × High-quality AI data set = Development of "Practical AI"

  1. HOME
  2. AI-Solution

Global Walkers provides you with a wide variety of AI solutions by multiplying our AI training data creation service called “Annotation One” by its development capability for Deep Learning.
Our development results include a broad range of solutions including person/object detection for security/surveillance camera, defective product detection for FA (Factory Automation) and crack/rust detection for building maintenance.
In the midst of the AI boom, there are very few AI systems which have been put to practical use.
Global Walkers provides AI solutions at a practical level by combining creation of high-quality training data, computer vision/image processing/3-D measurement, real-time AI framework and know-how to select hardware, in addition to know-how on Deep Learning.

Feature of AI-Solution

Deep Learning

Deep Learning

We are developing various models for Deep Learning including AI-OCR (character recognition), GW-Pose (3-D posture estimation), object detection model and area segmentation model. For instance, we have development results at a practical level such as person/object detection for security cameras.

Training Data Making

Training Data Making

We efficiently create high-quality training data by using self-developed “Data Augmentation Library” in addition to our training data creation service “Annotation One”. We maximize the performance of AI by increasing the training data by background synthesis/composition in addition to the basic Data Augmentation such as rotation or scaling.

Hardware

Hardware

We have expertise about hardware such as image sensing devices, lenses, filters and small computers. We provide you with total support ranging from selection of hardware to implementation of software.

Real-Time Processing

Real-Time Processing

Image capturing, image recognition processing by AI, posture estimation processing by AI and use of “GW-Edge”, a framework to process notice to servers in real time, enable efficient development.

Image Processing/3-D measurement

We realize AI systems at a practical level by combining moving object detection and tracking, etc.
Furthermore, we measure 3-D information which is unobtainable only from 2-D images by making full use of our know-how on camera calibration based on development of in-vehicle stereo cameras, stereo 3-D measurement, multi-view geometry and 3-D measurement using Depth cameras. We have achievements such as orbit/track of human head/ball by using 4 cameras.

Introduction results

Character Detection

Character Detection

Human Detection

Human Detection

Object Detection

Object Detection

Many achievements such as license plate detection system for coin parking, human / object detection system for surveillance robots, etc.

Development Flows

  1. Step1

    Requirements Definition

    We establish requirements for AI systems based on the issues you wish to solve. We support you as a technical consultant for selection of hardware to acquire data for entry, selection of computers for processing and selection of AI models.

  2. Step2

    PoC

    We conduct a trial called “PoC” (Proof of Concept) to confirm the feasibility of AI systems.

  3. Step3

    Prototype Development

    We develop a prototype after the PoC stage is cleared. We proceed with data collection, annotation, development/learning of AI models and GUI development. We build systems at the level to enable us to verify movements or to conduct demonstrations.

  4. Step4

    Development

    We extract scenes of AI’s misrecognition/unrecognition, create additional training data and perform additional learning of AI models. Moreover, we develop and test the entire systems and cause the AI systems to work at a practical level.

  5. Step5

    Additional Learning

    For maintenance and improvement of accuracy, after (in particular, immediately after) starting the system operation, we periodically conduct extraction of scenes of AI’s misrecognition/unrecognition, creation of additional training data and additional learning of AI models. We efficiently realize an improvement of accuracy of the AI models by using HITL-type data creation by “Annotation One” which is a data set creation service.

Technologies

Contact

FAQ

We have no knowledge about AI development. Can we make a request for it to you?

Yes, we can accept your request.
After hearing your requests or challenges, we will make a proposal to you on Necessary processes and development plans, etc. based on them.
There is no problem with your request level of "I want to do it this way.", so please feel free to contact us first.

Can we make a request for AI development for any purpose other than images, such as voice recognition?

Yes, we can accept your request.
We can widely handle voice-to-text processing, natural language processing or Anomaly detection based on signals from various sensors, so please feel free to contact us.