Abstract
A study on object detection utilizing deep learning is in continuous progress to promptly and accurately determine the surrounding situation in the driving environment. Existing studies have tried to improve object detection performance considering occlusion through various processes. However, recent studies use R-CNN-based deep learning to provide high accuracy at slow speeds, so there are limitations to real-time. In addition, since such previous studies never took into consideration the data imbalance problem of the objects of interest in the model training process, it is necessary to make additional improvements. Accordingly, we proposed a detection model of occluded object based on YOLO using hard-example mining and augmentation policy optimization. The proposed procedures were as follows: diverse augmentation policies were applied to the base model in sequence and the optimized policy suitable for training data were strategically selected through the gradient-based performance improvement rate. Then, in the model learning process, the occluded objects and the objects likely to induce a false-positive detection were extracted, and fine-tuning using transfer learning was conducted. As a result of the performance evaluation, the model proposed in this study showed an mAP@0.5 value of 90.49% and an F1-score value of 90%. It showed that this model detected occluded objects more stably and significantly enhanced the self-driving object detection accuracy compared with existing model.
Highlights
As the limitations of the means of transportation are significantly decreased due to the progress achieved by road traffic technology, the number of modern people who own a vehicle continues to increase, and this leads to diverse traffic problems
A detection model of occluded object based on YOLO using hard-example mining and augmentation policy optimization was proposed to detect objects that occlude a crosswalk on a real-time basis
Fine-tuning that applies transfer learning to the pre-trained weights was conducted, and a robust detection model suitable for hard examples was developed
Summary
As the limitations of the means of transportation are significantly decreased due to the progress achieved by road traffic technology, the number of modern people who own a vehicle continues to increase, and this leads to diverse traffic problems. The bounding box information of the classes defined based on the extracted vectors were taken into consideration, the classifier was re-designed to classify the hard-negative examples as a separate class from the ground truth, and the involved examples were included into the new learning category This induces the hard-negative objects likely to be detected as positive classes in the actual crosswalk detection process to be classified as classes of another category. The output value calculated through the hard-example mining algorithm represents the hardnegative examples and hard-positive examples extracted from the pre-existing positive examples and negative examples Since this serves as a factor that increases the falsepositive and false-negative detection rates in the process of detecting the objects of interest, the class bounding box of the involved data was reset and added to the pre-existing transfer learning model, and the data were re-learned.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
