We investigated the influence of horizontal shifts of the input images for one stage object detection method. We found that the object detector class scores drop when the target object center is at the grid boundary. Many approaches have focused on reducing the aliasing effect of down-sampling to achieve shift-invariance. However, down-sampling does not completely solve this problem at the grid boundary; it is necessary to suppress the dispersion of features in pixels close to the grid boundary into adjacent grid cells. Therefore, this paper proposes two approaches focused on the grid boundary to improve this weak point of current object detection methods. One is the Sub-Grid Feature Extraction Module, in which the sub-grid features are added to the input of the classification head. The other is Grid-Aware Data Augmentation, where augmented data are generated by the grid-level shifts and are used in training. The effectiveness of the proposed approaches is demonstrated using the COCO validation set after applying the proposed method to the FCOS architecture.
Shinji UCHINOURA
from Hiroshima University
Takio KURITA
from Hiroshima University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Shinji UCHINOURA, Takio KURITA, "Improved Head and Data Augmentation to Reduce Artifacts at Grid Boundaries in Object Detection" in IEICE TRANSACTIONS on Information,
vol. E107-D, no. 1, pp. 115-124, January 2024, doi: 10.1587/transinf.2023EDP7079.
Abstract: We investigated the influence of horizontal shifts of the input images for one stage object detection method. We found that the object detector class scores drop when the target object center is at the grid boundary. Many approaches have focused on reducing the aliasing effect of down-sampling to achieve shift-invariance. However, down-sampling does not completely solve this problem at the grid boundary; it is necessary to suppress the dispersion of features in pixels close to the grid boundary into adjacent grid cells. Therefore, this paper proposes two approaches focused on the grid boundary to improve this weak point of current object detection methods. One is the Sub-Grid Feature Extraction Module, in which the sub-grid features are added to the input of the classification head. The other is Grid-Aware Data Augmentation, where augmented data are generated by the grid-level shifts and are used in training. The effectiveness of the proposed approaches is demonstrated using the COCO validation set after applying the proposed method to the FCOS architecture.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2023EDP7079/_p
Copy
@ARTICLE{e107-d_1_115,
author={Shinji UCHINOURA, Takio KURITA, },
journal={IEICE TRANSACTIONS on Information},
title={Improved Head and Data Augmentation to Reduce Artifacts at Grid Boundaries in Object Detection},
year={2024},
volume={E107-D},
number={1},
pages={115-124},
abstract={We investigated the influence of horizontal shifts of the input images for one stage object detection method. We found that the object detector class scores drop when the target object center is at the grid boundary. Many approaches have focused on reducing the aliasing effect of down-sampling to achieve shift-invariance. However, down-sampling does not completely solve this problem at the grid boundary; it is necessary to suppress the dispersion of features in pixels close to the grid boundary into adjacent grid cells. Therefore, this paper proposes two approaches focused on the grid boundary to improve this weak point of current object detection methods. One is the Sub-Grid Feature Extraction Module, in which the sub-grid features are added to the input of the classification head. The other is Grid-Aware Data Augmentation, where augmented data are generated by the grid-level shifts and are used in training. The effectiveness of the proposed approaches is demonstrated using the COCO validation set after applying the proposed method to the FCOS architecture.},
keywords={},
doi={10.1587/transinf.2023EDP7079},
ISSN={1745-1361},
month={January},}
Copy
TY - JOUR
TI - Improved Head and Data Augmentation to Reduce Artifacts at Grid Boundaries in Object Detection
T2 - IEICE TRANSACTIONS on Information
SP - 115
EP - 124
AU - Shinji UCHINOURA
AU - Takio KURITA
PY - 2024
DO - 10.1587/transinf.2023EDP7079
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E107-D
IS - 1
JA - IEICE TRANSACTIONS on Information
Y1 - January 2024
AB - We investigated the influence of horizontal shifts of the input images for one stage object detection method. We found that the object detector class scores drop when the target object center is at the grid boundary. Many approaches have focused on reducing the aliasing effect of down-sampling to achieve shift-invariance. However, down-sampling does not completely solve this problem at the grid boundary; it is necessary to suppress the dispersion of features in pixels close to the grid boundary into adjacent grid cells. Therefore, this paper proposes two approaches focused on the grid boundary to improve this weak point of current object detection methods. One is the Sub-Grid Feature Extraction Module, in which the sub-grid features are added to the input of the classification head. The other is Grid-Aware Data Augmentation, where augmented data are generated by the grid-level shifts and are used in training. The effectiveness of the proposed approaches is demonstrated using the COCO validation set after applying the proposed method to the FCOS architecture.
ER -