Optimizing Intersection-Over-Union in Deep Neural Networks for Image Segmentation

By Md.Atiqur Rahman, Yang Wang. 2016

950 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression (2019, methodology): “Optimizing IoU using an approximate or a surrogate function: In the semantic segmentation task, there have been some efforts to optimize IoU using either an approximate function [18] or a surrogate loss [17].”
Reviving Iterative Training with Mask Guidance for Interactive Segmentation (2021, methodology): “We compare four loss functions that were used for training segmentation methods in recent works: binary cross entropy (BCE) loss, focal loss (FL) [37], soft IoU loss [51] and normalized focal loss (NFL) [38].”
Optimizing the Dice Score and Jaccard Index for Medical Image Segmentation: Theory and Practice (2019, neutral): “soft Jaccard [17] and Lovász-softmax [5]).”
The Lov´asz-Softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks (methodology): “As seen in Table A.2 the proxy loss in [23] does not reach the performance of our method.”
Learning to Balance Specificity and Invariance for In and Out of Domain Generalization (2020, methodology): “Therefore, inspired by prior work [40], we minimize the following soft-overlap loss for every pair of source domain masks {mi ,mj} at a layer L as,”
Attention-Guided Pyramid Context Networks for Detecting Infrared Small Target Under Complex Background (2023, methodology): “Batch size is set to 8, the loss function of each model is SoftIoU [37] and is trained for 5 epochs.”
MASK-RL: Multiagent Video Object Segmentation Framework Through Reinforcement Learning (2020, background): “…are given in Table VI, showing that the dice loss leads to better performance than cross-entropy loss (as already found in recent works, e.g., [34] and [40], especially in cases of unbalanced data sets) and that increasing the number of input frames does not affect performance signiﬁcantly—at…”
Alpha-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression (2021, background): “, Intersection over Union (IoU), as the localization loss [28, 39].”
BASNet: Boundary-Aware Salient Object Detection (2019, methodology): “Recently, it has been used as the training loss [56, 42].”
Asymmetric Contextual Modulation for Infrared Small Target Detection (2020, methodology): “We choose the SoftIoU [26] as the loss function and the Nesterov Accelerated Gradient method as the optimizer.”
Dynamic Task Prioritization for Multitask Learning (2018, methodology): “In the case where the KPI t is differentiable, such as the intersection-over-union loss layer [71] or KPI approximations [72], the gradient can be computed as normal.”
The Segment Anything Model (SAM) for Remote Sensing Applications: From Zero to One Shot (2023, background): “It measures the overlap between the predicted segmentation and the ground truth [45].”
PseudoClick: Interactive Image Segmentation with Click Imitation (2022, methodology): “In this study, we train PseudoClick models using four different loss functions: binary cross entropy (BCE) loss and focal loss (FL) [37], Soft IoU loss [54], and normalized focal loss (NFL) [36].”
F3Net: Fusion, Feedback and Focus for Salient Object Detection (2019, methodology): “IoU loss has been widely used in image segmentation (Rahman and Wang 2016).”
A General Survey on Attention Mechanisms in Deep Learning (2022, methodology): “For computer vision tasks, the PSNR [112], SSIM [113], or IoU [114] metrics are used.”
Attentional Local Contrast Networks for Infrared Small Target Detection (2020, background): “tion [40] for this highly unbalanced segmentation task, which is defined as follows:”