Batching Soft IoU for Training Semantic Segmentation Networks

Huang, Yifeng; Tang, Zhirong; Chen, Dan; Su, Kaixiong; Chen, Chengbin

doi:10.1109/lsp.2019.2956367

Cited by 35 publications

(11 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To evaluate the model's performance during the training process and the inference, we used two common metrics, namely the F1 score (Nava et al, 2022) and the Intersection over Union (IOU) score (e.g., Huang et al, 2019). We did not use binary accuracy because it is heavily influenced by data imbalance (Li et al, 2022) and can produce high accuracy, even for poor classifications.…”

Section: Susceptibility Componentmentioning

confidence: 99%

Space-time landslide hazard modeling via Ensemble Neural Networks

Dahal¹,

Tanyaş²,

Westen³

et al. 2022

Preprint

View full text Add to dashboard Cite

For decades, a full numerical description of the spatio-temporal dynamics of a landslide could be achieved only via physics-based models. The part of the geomorphology community focusing on data-driven model has instead focused on predicting where landslides may occur via susceptibility models. Moreover, they have estimated when landslides may occur via models that belong to the early-warning-system or to the rainfall-threshold themes. In this context, few published research have explored a joint spatio-temporal model structure. Furthermore, the third element completing the hazard definition, i.e., the landslide size, has hardly ever been modeled over space and time. However, the technological advancements of data-driven models have reached a level of maturity that allows to model all three components (Where, When and Size) mentioned above. This work, takes this direction and proposes for the first time a solution to the assessment of landslide hazard in a given area by jointly modeling landslide occurrences and their associated areal density per mapping unit, in space and time. To achieve this ambitious task, we have used a spatio-temporal landslide database generated for the Nepalese region affected by the Gorkha earthquake on the 25th of April 2015. The model relies on a deep-learning architecture trained using an Ensemble Neural Network, where the landslide occurrences and densities are aggregated over a squared mapping unit of 1 x 1 km and classified/regressed against a nested 30 m lattice. At the nested level, we have expressed predisposing and triggering factors. As for the temporal units, we have used an approximately 6-month resolution depending on the mapped inventory dates. The results are promising as our model performs satisfactorily both in the classification (susceptibility) and regression (density prediction) tasks. We believe that the model we propose brings a level of novelty that has the potential to create a rift with respect to the common susceptibility literature, finally proposing an integrated framework for hazard modeling in a data-driven context.

show abstract

Section: Susceptibility Componentmentioning

confidence: 99%

Space-time landslide hazard modeling via Ensemble Neural Networks

Dahal¹,

Tanyaş²,

Westen³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…To optimize the model, we combine L DM I and L IOU (Huang et al 2019) to minimize the loss function L for current epoch:…”

Section: Unsupervised Iteration Strategymentioning

confidence: 99%

Deep Neural Networks Learn Meta-Structures from Noisy Labels in Semantic Segmentation

Luo

Liu

Guo

et al. 2022

AAAI

View full text Add to dashboard Cite

How deep neural networks (DNNs) learn from noisy labels has been studied extensively in image classification but much less in image segmentation. So far, our understanding of the learning behavior of DNNs trained by noisy segmentation labels remains limited. In this study, we address this deficiency in both binary segmentation of biological microscopy images and multi-class segmentation of natural images. We generate extremely noisy labels by randomly sampling a small fraction (e.g., 10%) or flipping a large fraction (e.g., 90%) of the ground truth labels. When trained with these noisy labels, DNNs provide largely the same segmentation performance as trained by the original ground truth. This indicates that DNNs learn structures hidden in labels rather than pixel-level labels per se in their supervised training for semantic segmentation. We refer to these hidden structures in labels as meta-structures. When DNNs are trained by labels with different perturbations to the meta-structure, we find consistent degradation in their segmentation performance. In contrast, incorporation of meta-structure information substantially improves performance of an unsupervised segmentation model developed for binary semantic segmentation. We define meta-structures mathematically as spatial density distributions and show both theoretically and experimentally how this formulation explains key observed learning behavior of DNNs.

show abstract

“…However, Focal loss causes more false alarm due to the high response in the background suspicious area. SoftIoU loss [42] focuses on large-scale targets and lose small-scale targets. That is because, large-scale targets contributes much more in intersection over union (IoU ) than small-scale targets, resulting in lost of small-scale targets.…”

Section: F Focaliou Lossmentioning

confidence: 99%

MTU-Net: Multi-level TransUNet for Space-based Infrared Tiny Ship Detection

Wu¹,

Li²,

Luo³

et al. 2022

Preprint

View full text Add to dashboard Cite

Space-based infrared tiny ship detection aims at separating tiny ships from the images captured by earth orbiting satellites. Due to the extremely large image coverage area (e.g., thousands square kilometers), candidate targets in these images are much smaller, dimer, more changeable than those targets observed by aerial-based and land-based imaging devices. Existing short imaging distance-based infrared datasets and target detection methods cannot be well adopted to the space-based surveillance task. To address these problems, we develop a spacebased infrared tiny ship detection dataset (namely, NUDT-SIRST-Sea) with 48 space-based infrared images and 17598 pixel-level tiny ship annotations. Each image covers about 10000 square kilometers of area with 10000 × 10000 pixels. Considering the extreme characteristics (e.g., small, dim, changeable) of those tiny ships in such challenging scenes, we propose a multi-level TransUNet (MTU-Net) in this paper. Specifically, we design a Vision Transformer (ViT) Convolutional Neural Network (CNN) hybrid encoder to extract multi-level features. Local feature maps are first extracted by several convolution layers and then fed into the multi-level feature extraction module (MVTM) to capture long-distance dependency. We further propose a copy-rotateresize-paste (CRRP) data augmentation approach to accelerate the training phase, which effectively alleviates the issue of sample imbalance between targets and background. Besides, we design a FocalIoU loss to achieve both target localization and shape description. Experimental results on the NUDT-SIRST-Sea dataset show that our MTU-Net outperforms traditional and existing deep learning based SIRST methods in terms of probability of detection, false alarm rate and intersection over union. Our code is available at: https://github.com/TianhaoWu16/Multi-level-TransUNet-for-Space-based-Infrared-Tiny-ship-Detection.

show abstract

Batching Soft IoU for Training Semantic Segmentation Networks

Cited by 35 publications

References 18 publications

Space-time landslide hazard modeling via Ensemble Neural Networks

Space-time landslide hazard modeling via Ensemble Neural Networks

Deep Neural Networks Learn Meta-Structures from Noisy Labels in Semantic Segmentation

MTU-Net: Multi-level TransUNet for Space-based Infrared Tiny Ship Detection

Contact Info

Product

Resources

About