Field Application

Auto Deep Learning
MLOps pipeline

Build on-site MLOps from initial model training
to deployment and maintenance.

Challenges faced during
retraining on-site

Retraining models with newly acquired data and deploying them back to the production line takes a significant amount of time.

Solution

Auto Deep Learning MLOps

Solution

Auto Deep Learning MLOps

Scenarios

Neurocle MLOps scenarios

Scenarios

Neurocle MLOps scenarios

MLOps

1. On-device MLOps

Perform real-time inspection and retraining on on-site equipment simultaneously.

2. Centralized Server MLOps

Transmit data from on-site equipment to a central server for centralized training and model management.

MLOps

1. On-device MLOps

Perform real-time inspection and retraining on on-site equipment simultaneously.

2. Centralized Server MLOps

Transmit data from on-site equipment to a central server for centralized training and model management.

MLOps

1. On-device MLOps

Perform real-time inspection and retraining on on-site equipment simultaneously.

2. Centralized Server MLOps

Transmit data from on-site equipment to a central server for centralized training and model management.

Scenario 1

On-device MLOps

Manage model training, deployment and maintenance at the equipment level.

MLOps

Workflow 1

Fully automated workflow based on existing label information

Perform automatic labeling with information from existing models. Maintain accuracy by selectively utilizing only the predictions with 90% or higher confidence levels.

MLOps

Workflow 1

Fully automated workflow based on existing label information

Perform automatic labeling with information from existing models. Maintain accuracy by selectively utilizing only the predictions with 90% or higher confidence levels.

MLOps

Workflow 1

Fully automated workflow based on existing label information

Perform automatic labeling with information from existing models. Maintain accuracy by selectively utilizing only the predictions with 90% or higher confidence levels.

MLOps

Workflow 2

External labeling workflow with existing or third-party labeling tools

Design customized labeling tool, or label with existing equipment or third-party tools. Neuro-T Engine integrates flexibly without UI constraints and therefore allows use of any UI and tools as needed.

MLOps

Workflow 2

External labeling workflow with existing or third-party labeling tools

Design customized labeling tool, or label with existing equipment or third-party tools. Neuro-T Engine integrates flexibly without UI constraints and therefore allows use of any UI and tools as needed.

MLOps

Workflow 2

External labeling workflow with existing or third-party labeling tools

Design customized labeling tool, or label with existing equipment or third-party tools. Neuro-T Engine integrates flexibly without UI constraints and therefore allows use of any UI and tools as needed.

Scenario 2

Centralized Server MLOps

Manage model training and operation at a central server and maintain consistent labeling standards.

Collect new data and conduct labeling in the central server and distribute label information to each equipment.

The centralized server architecture offers the following benefits:
- Consistent labeling standards
- Efficient training in each equipment to avoid training bottlenecks
- Centralized monitoring training status across every equipment.

Product introduction

Looking for modularized software for fast MLOps implementation?

Auto Deep Learning Model Trainer

GUI-based training software for generating high-performance inspection models without deep learning expertise

View more

Auto Deep Learning Training Engine

High-performance Auto Deep Learning training module for flexible integration on various systems.

View more

Runtime Library for On-site Deployment

Runtime library for real-time, on-site deep learning insepction model operation

View more

Runtime Library for On-site Deployment

View more

Runtime Library for On-site Deployment

View more