| 000 | 02319nam a22002537a 4500 | ||
|---|---|---|---|
| 003 | OSt | ||
| 005 | 20231021150021.0 | ||
| 008 | 230822b |||||||| |||| 00| 0 eng d | ||
| 040 |
_aTUPM _beng _c- _erda |
||
| 050 |
_aDIS Q 325.73 _bR49 2022 |
||
| 100 | _aReyes, Ryan C. | ||
| 245 |
_aDeep learning based road damage detection and classification for Philippine road pavements _bRyan C. Reyes |
||
| 264 |
_aManila _b: Technological University of the Philippines _c2022 |
||
| 300 |
_a143 pages: _bcolor illustration; _c28cm. _e+1 CD-ROM (4 3/4 in.) |
||
| 500 | _aDissertation | ||
| 502 |
_aCollege of Industrial Education-- _bDoctor of Technology. _cTechnological University of the Philippines. _d2022 |
||
| 520 | 3 | _aThis Paper presents a YOLO based road damage detection and classification model that employs transfer learning approach where a pre-trained YOLOv7 model was trained for custom Philippine road damage datasets which are collections of 8004 images of Philippine road pavement containing instances of different road damages. in this supervised learning project, the bulk collection of images was labeled by the road experts for annotation using the on-line computer vision annotation tool (CVAT by Intel) to highlight the part of the images that contains the damage features fed in the learning training algorithm. the model development process begins with experimentations using initial datasets to determine the best model for the desk and proceeded with the final system of DL classifiers to predict the damage type of an input image. Experimentation verified the dominance of YOLOv7 which outperformed two other candidate YOLO variants for the task in terms of mean average precision and F1 score using initial datasets in the early stage of this study. the model was retrained with bigger datasets and produced a more competitive detection performance. the study also explored two (2) methods for the detection of road damage severity levels using the same datasets distributed across all severity labeled by road experts resulting to a slight decrease in the mean average precision scores and F1 measures-Author's Abstract | |
| 650 | _aDeep learning (Machine) | ||
| 650 | _aYOLO | ||
| 650 | _aMachine learning | ||
| 650 | _aMachine vision | ||
| 650 | _aRoad damage detection | ||
| 942 |
_2lcc _cDIS _n0 |
||
| 999 |
_c28078 _d28078 |
||