Abstract
Environment recognition has been an important topic ever since the emergence of augmented reality (AR). For better experience in AR applications, environment recognition should be provided fast in real-time, where real-time object detection technologies could fulfill this requirement. However, training object detectors for AR specific scenarios are often troublesome. The real-time nature of AR produces visual degradations such as motion blur or occlusion by interaction, which make detectors trained with plain data difficult to detect objects exposed in such complex situations. Also, since gathering and labeling training data from scratch is a heavy burden, we need to resort to synthesized training data but previous synthetic data generation frameworks do not consider the aforementioned issue. Therefore, in this paper, we propose a new synthetic data generation framework which includes visual variations such as motion blur and occlusion occurred by distractors. By this simple modification, we show that including such variated data to the training dataset could dramatically improve realtime performance of object detectors by a high margin. Also, we stress that synthesizing training data with no more than three objects per image can achieve competitive performance compared to detectors trained with over four present in a single image. Experimental results both quantitatively and qualitatively supports our statements and shows the superiority of our method.
| Original language | English |
|---|---|
| Title of host publication | 2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Proceedings |
| Publisher | Institute of Electrical and Electronics Engineers Inc. |
| Pages | 1518-1522 |
| Number of pages | 5 |
| ISBN (Electronic) | 9789881476852 |
| DOIs | |
| Publication status | Published - 2019 Mar 4 |
| Event | 10th Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Honolulu, United States Duration: 2018 Nov 12 → 2018 Nov 15 |
Publication series
| Name | 2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Proceedings |
|---|
Conference
| Conference | 10th Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 |
|---|---|
| Country | United States |
| City | Honolulu |
| Period | 18/11/12 → 18/11/15 |
Fingerprint
All Science Journal Classification (ASJC) codes
- Information Systems
Cite this
}
A Simple Method on Generating Synthetic Data for Training Real-time Object Detection Networks. / Huh, Jungwoo; Lee, Kyoungoh; Lee, Inwoong; Lee, Sanghoon.
2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2019. p. 1518-1522 8659778 (2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Proceedings).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
TY - GEN
T1 - A Simple Method on Generating Synthetic Data for Training Real-time Object Detection Networks
AU - Huh, Jungwoo
AU - Lee, Kyoungoh
AU - Lee, Inwoong
AU - Lee, Sanghoon
PY - 2019/3/4
Y1 - 2019/3/4
N2 - Environment recognition has been an important topic ever since the emergence of augmented reality (AR). For better experience in AR applications, environment recognition should be provided fast in real-time, where real-time object detection technologies could fulfill this requirement. However, training object detectors for AR specific scenarios are often troublesome. The real-time nature of AR produces visual degradations such as motion blur or occlusion by interaction, which make detectors trained with plain data difficult to detect objects exposed in such complex situations. Also, since gathering and labeling training data from scratch is a heavy burden, we need to resort to synthesized training data but previous synthetic data generation frameworks do not consider the aforementioned issue. Therefore, in this paper, we propose a new synthetic data generation framework which includes visual variations such as motion blur and occlusion occurred by distractors. By this simple modification, we show that including such variated data to the training dataset could dramatically improve realtime performance of object detectors by a high margin. Also, we stress that synthesizing training data with no more than three objects per image can achieve competitive performance compared to detectors trained with over four present in a single image. Experimental results both quantitatively and qualitatively supports our statements and shows the superiority of our method.
AB - Environment recognition has been an important topic ever since the emergence of augmented reality (AR). For better experience in AR applications, environment recognition should be provided fast in real-time, where real-time object detection technologies could fulfill this requirement. However, training object detectors for AR specific scenarios are often troublesome. The real-time nature of AR produces visual degradations such as motion blur or occlusion by interaction, which make detectors trained with plain data difficult to detect objects exposed in such complex situations. Also, since gathering and labeling training data from scratch is a heavy burden, we need to resort to synthesized training data but previous synthetic data generation frameworks do not consider the aforementioned issue. Therefore, in this paper, we propose a new synthetic data generation framework which includes visual variations such as motion blur and occlusion occurred by distractors. By this simple modification, we show that including such variated data to the training dataset could dramatically improve realtime performance of object detectors by a high margin. Also, we stress that synthesizing training data with no more than three objects per image can achieve competitive performance compared to detectors trained with over four present in a single image. Experimental results both quantitatively and qualitatively supports our statements and shows the superiority of our method.
UR - http://www.scopus.com/inward/record.url?scp=85063439166&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85063439166&partnerID=8YFLogxK
U2 - 10.23919/APSIPA.2018.8659778
DO - 10.23919/APSIPA.2018.8659778
M3 - Conference contribution
AN - SCOPUS:85063439166
T3 - 2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Proceedings
SP - 1518
EP - 1522
BT - 2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
ER -