skip to main content
Idioma:

Focal and efficient IOU loss for accurate bounding box regression

Zhang, Yi-Fan ; Ren, Weiqiang ; Zhang, Zhang ; Jia, Zhen ; Wang, Liang ; Tan, Tieniu

Neurocomputing (Amsterdam), 2022-09, Vol.506, p.146-157 [Revista revisada por pares]

Elsevier B.V

Texto completo disponible

Citas Citado por
  • Título:
    Focal and efficient IOU loss for accurate bounding box regression
  • Autor: Zhang, Yi-Fan ; Ren, Weiqiang ; Zhang, Zhang ; Jia, Zhen ; Wang, Liang ; Tan, Tieniu
  • Materias: Hard sample mining ; Loss function design ; Object detection
  • Es parte de: Neurocomputing (Amsterdam), 2022-09, Vol.506, p.146-157
  • Descripción: Illustrations on the problems of current BBR losses. Each row shows the optimization results in different iterations with certain loss function. The Black denotes the anchor box. The Blue denotes the target box. The fist row denotes GIOU. The second row denotes CIOU. The third row denotes the proposed EIOU. EIOU attains more quick convergence speed and more accurate regression results. [Display omitted] •We reveal the flaws of ℓn-norm and IOU-based losses for object detection.•We design a regression version of focal loss to emphasize the most promising anchors.•We conduct extensive experiments to validate the superiority of the proposed methods. In object detection, bounding box regression (BBR) is a crucial step that determines the object localization performance. However, we find that most previous loss functions for BBR have two main drawbacks: (i) Both ℓn-norm and IOU-based loss functions are inefficient to depict the objective of BBR, which leads to slow convergence and inaccurate regression results. (ii) Most of the loss functions ignore the imbalance problem in BBR that the large number of anchor boxes which have small overlaps with the target boxes contribute most to the optimization of BBR. To mitigate the adverse effects caused thereby, we perform thorough studies to exploit the potential of BBR losses in this paper. Firstly, an Efficient Intersection over Union (EIOU) loss is proposed, which explicitly measures the discrepancies of three geometric factors in BBR, i.e., the overlap area, the central point and the side length. After that, we state the Effective Example Mining (EEM) problem and propose a regression version of focal loss to make the regression process focus on high-quality anchor boxes. Finally, the above two parts are combined to obtain a new loss function, namely Focal-EIOU loss. Extensive experiments on both synthetic and real datasets are performed. Notable superiorities on both the convergence speed and the localization accuracy can be achieved over other BBR losses.
  • Editor: Elsevier B.V
  • Idioma: Inglés
Volver a la lista de resultados
Ir a la página anteriorAnterior Resultado  2 SiguienteIr a la siguiente página

  • Realización: ABCD-USP USP   Logos de Redes Sociales: Freepik

Buscando en bases de datos remotas, por favor espere

  • Buscando por
  • enscope:(USP_VIDEOS),scope:("PRIMO"),scope:(USP_FISICO),scope:(USP_EREVISTAS),scope:(USP),scope:(USP_EBOOKS),scope:(USP_PRODUCAO),primo_central_multiple_fe
  • Mostrar lo que tiene hasta ahora