In this paper, we propose a Label Reassignment Strategy to improve the performance of an object detection algorithm. Our approach involves two stages: an inference stage and an assignment stage. In the inference stage, we perform multi-scale inference with predefined scale sizes on a trained model and re-infer masked images to obtain robust classification results. In the assignment stage, we calculate the IoU between bounding boxes to remove duplicates. We also check box and class occurrence between the detection result and annotation label to re-assign the dominant class type. We trained the YOLOX-L model with the re-annotated dataset to validate our strategy. The model achieved a 3.9% improvement in mAP and 3x better performance on AP_S compared to the model trained with the original dataset. Our results demonstrate that the proposed Label Reassignment Strategy can effectively improve the performance of an object detection model.
@article{ART002908028}, author={Jung-In An and Yoon Kim and Hyun-Soo Choi}, title={Small-Scale Object Detection Label Reassignment Strategy}, journal={Journal of The Korea Society of Computer and Information}, issn={1598-849X}, year={2022}, volume={27}, number={12}, pages={77-84}, doi={10.9708/jksci.2022.27.12.077}
TY - JOUR AU - Jung-In An AU - Yoon Kim AU - Hyun-Soo Choi TI - Small-Scale Object Detection Label Reassignment Strategy JO - Journal of The Korea Society of Computer and Information PY - 2022 VL - 27 IS - 12 PB - The Korean Society Of Computer And Information SP - 77 EP - 84 SN - 1598-849X AB - In this paper, we propose a Label Reassignment Strategy to improve the performance of an object detection algorithm. Our approach involves two stages: an inference stage and an assignment stage. In the inference stage, we perform multi-scale inference with predefined scale sizes on a trained model and re-infer masked images to obtain robust classification results. In the assignment stage, we calculate the IoU between bounding boxes to remove duplicates. We also check box and class occurrence between the detection result and annotation label to re-assign the dominant class type. We trained the YOLOX-L model with the re-annotated dataset to validate our strategy. The model achieved a 3.9% improvement in mAP and 3x better performance on AP_S compared to the model trained with the original dataset. Our results demonstrate that the proposed Label Reassignment Strategy can effectively improve the performance of an object detection model. KW - Artificial Intelligence;Object Detection;Dataset Refinement;Dataset Reassignment DO - 10.9708/jksci.2022.27.12.077 ER -
Jung-In An, Yoon Kim and Hyun-Soo Choi. (2022). Small-Scale Object Detection Label Reassignment Strategy. Journal of The Korea Society of Computer and Information, 27(12), 77-84.
Jung-In An, Yoon Kim and Hyun-Soo Choi. 2022, "Small-Scale Object Detection Label Reassignment Strategy", Journal of The Korea Society of Computer and Information, vol.27, no.12 pp.77-84. Available from: doi:10.9708/jksci.2022.27.12.077
Jung-In An, Yoon Kim, Hyun-Soo Choi "Small-Scale Object Detection Label Reassignment Strategy" Journal of The Korea Society of Computer and Information 27.12 pp.77-84 (2022) : 77.
Jung-In An, Yoon Kim, Hyun-Soo Choi. Small-Scale Object Detection Label Reassignment Strategy. 2022; 27(12), 77-84. Available from: doi:10.9708/jksci.2022.27.12.077
Jung-In An, Yoon Kim and Hyun-Soo Choi. "Small-Scale Object Detection Label Reassignment Strategy" Journal of The Korea Society of Computer and Information 27, no.12 (2022) : 77-84.doi: 10.9708/jksci.2022.27.12.077
Jung-In An; Yoon Kim; Hyun-Soo Choi. Small-Scale Object Detection Label Reassignment Strategy. Journal of The Korea Society of Computer and Information, 27(12), 77-84. doi: 10.9708/jksci.2022.27.12.077
Jung-In An; Yoon Kim; Hyun-Soo Choi. Small-Scale Object Detection Label Reassignment Strategy. Journal of The Korea Society of Computer and Information. 2022; 27(12) 77-84. doi: 10.9708/jksci.2022.27.12.077
Jung-In An, Yoon Kim, Hyun-Soo Choi. Small-Scale Object Detection Label Reassignment Strategy. 2022; 27(12), 77-84. Available from: doi:10.9708/jksci.2022.27.12.077
Jung-In An, Yoon Kim and Hyun-Soo Choi. "Small-Scale Object Detection Label Reassignment Strategy" Journal of The Korea Society of Computer and Information 27, no.12 (2022) : 77-84.doi: 10.9708/jksci.2022.27.12.077