Towrards Automation of Train Data Annotation for Object Detection Model
- Delivery
- Available on this site
- Format
- Price
- Non-members (tax incl.):¥1,100 Members (tax incl.):¥880
- Publication code
- 20215271
- Paper/Info type
- Proceedings (Spring)
No.61-21
- Pages
- 1-5(Total 5 p)
- Date of publication
- May 2021
- Publisher
- JSAE
- Language
- English
- Event
- 2021 JSAE Annual Congress (Spring)[Online Meeting]
Detailed Information
Author(E) | 1) Mohd Hafiz Hilman Mohammad Sofian, 2) Toshio Ito, 3) Yasutaka Okada, 4) Rei Hiromi, 5) Aoi Ogishima |
---|---|
Affiliation(E) | 1) Shibaura Institute of Technology, 2) Shibaura Institute of Technology, 3) DENSO TEN, 4) DENSO TEN, 5) DENSO TEN |
Abstract(E) | Convolutional Neural Network has been one of the main reasons for the development of the autonomous driving system. It may be used to detect the presence of other vehicles and pedestrians and the driver can be then alarmed for any possible accidents. However, CNN needs to be trained on a large annotated dataset which can introduce a high cost of time and human labor. We propose a system called "Machine Teachers" to reduce those costs by automating the data annotation process. This system will then be used with the Active Learning framework for a faster and probably better training process. |