Summary

Project to label training data for AI in an autonomous driving system using 3D and 2D tools.

Scale

42 Members

Team

Annotators
Modifier
QA

Method

Bounding box

Work flow

1. Receive data from customers
2. Image labelling
3. Secondary inspection (Modifier)
4. Final review (QA）
5. Data delivery

Result

• ◉ Data Processed in 1 Month: 132,000 images
• ◉ Accuracy Rate: Over 98%

Summary

Project to develop a system for target control and optimization of repetitive, decentralized factory work.

Scale

25 Members

Team

Annotator
Inspector
QA

Method

Point
Pixel Annotation

Work flow

1. Receive Data from Customer
2. Confirm Specifications
3. Image-Labelling
4. Primary Inspection (Self-check)
5. Secondary Inspection (QA)
6. Data Delivery

Result

• ◉ Productivity: 160 Images per day
• ◉ Accuracy: Over 99% (including Dialect)

Summary

Computer-aided real-time transcription system to convert voice into text for the Vietnamese National Assembly and local government councils.

Scale

100 Members

Team

Annotator
Inspector
QA

Work flow

1. Annotation Tool Development
2. Audio Data Collection
3. Data Labeling
4. Machine Learning
5. Build with Machine Learning Model

Result

• ◉ Processed data: 2,000+ hours of audio
• ◉ Accuracy Rate: 95% (including Dialect)
• ◉ Reduced Workload: 60%

Summary

Putting equipment-checking forms (excel, templates, etc), used in the customer’s nuclear power plants, and bringing them online to report anytime and anywhere using tablets or iPhone.

Scale

32 Man-Months

Team

Annotators
QA
PM

Work flow

1. Design excel template
2. Edit excel interface
3. Set Cell Clusters
4. Import to Comas Designer Tool to check & install more features
5. Use Windows i-Reporter to check the Actual Interface
6. Export the completed excel file
7. Review the results
8. Data Delivery

Result

• ◉ Data Processed in 1 Month: 350
• ◉ Accuracy: Over 90%