Improving 3D Monocular Object Detection with Dynamic Loss Function Adjustments
| dc.contributor.author | Evain, Alexandre | |
| dc.contributor.author | Khemmar, Redouane | |
| dc.contributor.author | Orzalesi, Mathieu | |
| dc.contributor.author | Ahmedali, Sofiane | |
| dc.contributor.editor | Skala, Václav | |
| dc.date.accessioned | 2025-07-30T09:45:29Z | |
| dc.date.available | 2025-07-30T09:45:29Z | |
| dc.date.issued | 2025 | |
| dc.description.abstract-translated | In 3D monocular object detection, optimizing the loss function is crucial for balancing multiple competing metrics, such as depth estimation, orientation, and object dimensions. Traditional approaches use a weighted sum of individual losses, allowing metric prioritization but risking training instability due to competition between terms. To address this, we first experimented with different loss function configurations to see how different loss interactions could emphasize specific metrics. These initial results demonstrated that abrupt changes in loss functions cause significant precision drops, therefore we decided to try dynamic loss functions adjustment, using transition functions to gradually shift metric emphasis over the training process. Among the tested transition functions, the Smoothstep function had the best balance across all metrics, followed by the Linear function, while the Smootherstep function provided strong initial performance but was eventually outperformed. Our results suggest that controlled, smooth transitions between different loss functions can enhance training stability and final detection accuracy, providing a way to improve 3D object detection models without overhauling their architecture | en |
| dc.description.sponsorship | This research is funded and supported by SEGULA Technologies. We would like to thank SEGULA Technologies for their collaboration and for allowing us to conduct this research. We would like to thank also the engineers of the Autonomous Navigation Laboratory (ANL) of IRSEEM for their support. In addition, this work was performed, in part, on computing resources provided by CRIANN (Centre Regional Informatique et d’Applications Numeriques de Normandie, Normandy, France). | |
| dc.format | 8 s. | cs |
| dc.format.mimetype | application/pdf | |
| dc.identifier.doi | http://www.doi.org/10.24132/CSRN.2025-17 | |
| dc.identifier.issn | 2464-4617 (Print) | |
| dc.identifier.issn | 2464-4625 (online) | |
| dc.identifier.uri | http://hdl.handle.net/11025/62223 | |
| dc.language.iso | en | en |
| dc.publisher | Vaclav Skala - UNION Agency | en |
| dc.rights | © Vaclav Skala - UNION Agency | en |
| dc.rights.access | openAccess | en |
| dc.subject | monokulární 3D detekce objektů | cs |
| dc.subject | optimalizace ztrátové funkce | cs |
| dc.subject | dynamické vyrovnávání ztrát | cs |
| dc.subject | přechodové funkce | cs |
| dc.subject | hluboké učení | cs |
| dc.subject | počítačové vidění | cs |
| dc.subject | trénovací stabilita | cs |
| dc.subject.translated | monocular 3D object detection | en |
| dc.subject.translated | loss function optimization | en |
| dc.subject.translated | dynamic loss adjustment | en |
| dc.subject.translated | transition functions | en |
| dc.subject.translated | deep learning | en |
| dc.subject.translated | computer vision | en |
| dc.subject.translated | training stability | en |
| dc.title | Improving 3D Monocular Object Detection with Dynamic Loss Function Adjustments | en |
| dc.type | konferenční příspěvek | cs |
| dc.type | conferenceObject | en |
| dc.type.status | Peer reviewed | en |
| dc.type.version | publishedVersion | en |
| local.files.count | 1 | * |
| local.files.size | 2193575 | * |
| local.has.files | yes | * |