Segmentation Loss Research Articles

Weight-guided loss for long-tailed object detection and instance segmentation

The long-tailed characteristic leads to a significant performance drop for various models on long-tailed distribution datasets. Existing works mainly focus on mitigating the data shortage in tail classes at dataset level by data re-sampling, loss re-weighting or knowledge transfer from head to tail. While in this paper, we focus on another perspective which is also related to the performance drop: the gap between total dataset class number and training batch size. To address this issue, we propose a Weight-Guided (WG) loss which utilizes the classifier weights as auxiliary tail samples. It can be easily deployed to different models. By simply adding WG loss to Mask R-CNN with ResNet-50 backbone, we improve the performance by (i) 0.5 box AP and 0.4 mask AP on COCO dataset, (ii) 0.4 box and mask AP (1.8 mask AP for rare classes) on LVIS v1.0 dataset. Codes will be released.

ELCD: Efficient Lunar Crater Detection Based on Attention Mechanisms and Multiscale Feature Fusion Networks from Digital Elevation Models

The detection and counting of lunar impact craters are crucial for the selection of detector landing sites and the estimation of the age of the Moon. However, traditional crater detection methods are based on machine learning and image processing technologies. These are inefficient for situations with different distributions, overlaps, and crater sizes, and most of them mainly focus on the accuracy of detection and ignore the efficiency. In this paper, we propose an efficient lunar crater detection (ELCD) algorithm based on a novel crater edge segmentation network (AFNet) to detect lunar craters from digital elevation model (DEM) data. First, in AFNet, a lightweight attention mechanism module is introduced to enhance the feature extract capabilities of networks, and a new multiscale feature fusion module is designed by fusing different multi-level feature maps to reduce the information loss of the output map. Then, considering the imbalance in the classification and the distributions of the crater data, an efficient crater edge segmentation loss function (CESL) is designed to improve the network optimization performance. Lastly, the crater positions are obtained from the network output map by the crater edge extraction (CEA) algorithm. The experiment was conducted on the PyTorch platform using two lunar crater catalogs to evaluate the ELCD. The experimental results show that ELCD has a superior detection accuracy and inference speed compared with other state-of-the-art crater detection algorithms. As with most crater detection models that use DEM data, some small craters may be considered to be noise that cannot be detected. The proposed algorithm can be used to improve the accuracy and speed of deep space probes in detecting candidate landing sites, and the discovery of new craters can increase the size of the original data set.

Research on CT Lung Segmentation Method of Preschool Children based on Traditional Image Processing and ResUnet.

Lung segmentation using computed tomography (CT) images is important for diagnosing various lung diseases. Currently, no lung segmentation method has been developed for assessing the CT images of preschool children, which may differ from those of adults due to (1) presence of artifacts caused by the shaking of children, (2) loss of a localized lung area due to a failure to hold their breath, and (3) a smaller CT chest area, compared with adults. To solve these unique problems, this study developed an automatic lung segmentation method by combining traditional imaging methods with ResUnet using the CT images of 60 children, aged 0-6 years. First, the CT images were cropped and zoomed through ecological operations to concentrate the segmentation task on the chest area. Then, a ResUnet model was used to improve the loss for lung segmentation, and case-based connected domain operations were performed to filter the segmentation results and improve segmentation accuracy. The proposed method demonstrated promising segmentation results on a test set of 12 cases, with average accuracy, Dice, precision, and recall of 0.9479, 0.9678, 0.9711, and 0.9715, respectively, which achieved the best performance relative to the other six models. This study shows that the proposed method can achieve good segmentation results in CT of preschool children, laying a good foundation for the diagnosis of children's lung diseases.

Analysis of the Effects of Segmentation Networks and Loss Functions in Ischemic Stroke Lesion Segmentation

Stroke was the cause of one out of every six deaths from cerebrovascular disease in 2020. A stroke occurs in the United States (US) every 40 seconds. Every 3.5 minutes, people die of a stroke. More than total 795,000 stroke cases occur yearly in the US. This study aims to detect the ischemic stroke lesion that occurs in the brain. The Ischemic Stroke Lesion Segmentation (ISLES) 2017 data set, which includes 82 Magnetic Resonance images of 43 patients, was used. The UNet, Attention UNet, Residual UNet, Attention Residual UNet, and Residual UNet++ segmentation networks were tested. Moreover, Cross Entropy, Dice, IoU, Tversky, Focal Tversky, and their compound forms were analyzed. The IoU loss function tested on Attention UNet achieved the best performance with the dice score of 0.766, the IoU score of 0.621, the sensitivity of 0.730, the specificity of 0.997, the precision of 0.805, and the accuracy of 0.993.

Superposition-Based URLLC Traffic Scheduling in 5G and Beyond Wireless Networks

Ultra-Reliable and Low Latency Communications (URLLC) is one of the essential services in 5G networks and beyond. The coexistence of URLLC alongside other services, namely, enhanced Mobile BroadBand (eMBB) and massive Machine-Type Communications (mMTC), calls for developing spectrally efficient multiplexing techniques. In this work, we study the problem of scheduling URLLC traffic in a downlink system in the presence of eMBB traffic. Based on the proposed superposition/puncturing scheme, a resource allocation problem is formulated with the objective to minimize the rate loss of the eMBB service and URLLC packet segmentation loss while satisfying the eMBB and URLLC quality of service (QoS) constraints. The resulting problem is formulated as a mixed-integer non-linear program (MINLP) which is generally very hard to solve in polynomial time. Hence, we reformulate the problem as a one-to-one pairing problem and we derive its feasibility region as well as the optimal solutions for the power and spectral resource allocation. Subsequently, we propose a low complexity algorithm to support the many-to-many pairing. Simulation results show that the proposed algorithm achieves higher URLLC packet admission rate and lower rate loss for eMBB. For instance, the URLLC packet admission rate, unlike baseline methods, is shown to be preserved under the proposed method even at higher URLLC load. It is shown that at least 30% more URLLC users can be served without degrading their QoS, while keeping the impact on eMBB rate minimal. Detailed numerical evaluation is presented to quantify the benefits of the proposed method.

Densely Connected Squeeze‐and‐Excitation Convolutional Encoder‐Decoder Networks for Identifying Preferential Channels in Highly Heterogeneous Porous Media

AbstractIn highly heterogeneous formations, the preferential channels (PC) that usually exist can dramatically change the transport dynamics of a solute plume. Identifying PC is vital to delineate contaminant transport and conduct the environmental risk analysis. Conventional methods usually depend on the solution of the governing equations for flow and transport and struggle with high computational costs. In this paper, we transform PC identification into a particular semantic segmentation task for the first time and build a densely connected squeeze‐and‐excitation convolutional encoder‐decoder network. A training strategy combining segmentation loss and new regression loss (index distance loss) is used to improve the identification of PC locations. We evaluate the effectiveness and generalizability of the proposed deep learning model with conductivity fields with different variances. A new connectivity indicator (connectivity ratio) is proposed to describe the connectivity area of a field. The network trained by conductivity fields with larger variances (higher heterogeneity) produces better PC identification and has a stronger generalizability. A conductivity field with a large variance usually has small connected areas and induces tortuous and long PC. The small connected areas restrict the generation of multiple PCs. This makes the model learn discriminative features easily. And the long PC increase its length, which means an increasing of the number of positive pixels in training samples. Finally, we compare the model with the stream function and minimum resistance methods. The results show that the deep learning model is an efficient method and can provide more connectivity information than the other two methods.

Nested U-Net with Enhanced Attention Gate and Compound Loss for Semantic Segmentation of Brain Tumor from Multimodal MRI

Nested U-Net with Enhanced Attention Gate and Compound Loss for Semantic Segmentation of Brain Tumor from Multimodal MRI

Uncertainty teacher with dense focal loss for semi-supervised medical image segmentation

In medical scenarios, obtaining pixel-level annotations for medical images is expensive and time-consuming, even if considering its importance for automating segmentation tasks. Due to the scarcity of labels in the training phase, semi-supervised methods are widely applied for various medical tasks. To better utilize the unlabeled data, several works have explored the method of uncertainty estimation and exhibited huge success. Despite their impressive performance, we believe that the underlying information of the unlabeled data has been largely unexplored. Meanwhile, there is an extreme foreground–background class imbalance during the training phase of semantic segmentation, which may cause a vast number of easily classified samples to overwhelm the loss during training and lead to a model collapse. In this paper, we proposed uncertainty teacher with dense focal loss, a method that can take good advantage of unlabeled data simultaneously and address the class imbalance problem, based on Deep Co-Training. On one hand, the uncertainty teacher framework is presented to better utilize the unlabeled data by introducing a novel method to regularize uncertainty in the right direction, and the uncertainty is estimated by Monte Carlo Sampling. On the other hand, the dense focal loss is proposed to help solve the class imbalance problem between different classes of samples in medical image segmentation and effectively convert the multi-variate entropy into a multiple binary entropy. We implemented our method on three challenging public medical datasets and experimental results have shown desirable improvements to state-of-the-art.

Distinguishing foreground and background alignment for unsupervised domain adaptative semantic segmentation

Unsupervised domain adaptive semantic segmentation uses the knowledge learned from the labeled source domain dataset to guide the segmentation of the target domain. However, this domain migration method will cause a large inter-domain difference due to the different feature distributions between the source domain and the target domain. We use the self-supervised learning method to generate pseudo labels for the target domain, so that the corresponding pixels are directly aligned with the source domain according to the segmentation loss. Through observation, it is found that the spatial distribution of the background class in the source domain and the target domain has a small difference, and the appearance of the same class of the foreground class will also be quite different. We use the method of distinguishing alignment between foreground and background classes. We understand that acquiring the rich space and channel information on the feature map during the convolution process is essential for fine-grained semantic segmentation. Therefore, in order to obtain the dependency relationship between the channels of the feature map and the spatial position information, we use a channel and spatial parallel attention module. This module enables the network to select and amplify valuable space and channel information from the global information and suppress useless information. In addition, we introduce focal loss to solve the problem of class imbalance in the data set. Experiments show that our method achieves better segmentation performance in unsupervised domain adaptive semantic segmentation. • We use self-supervised learning to generate pseudo-labels for the target domain. • We distinguish and align the foreground and background classes. • We use parallel attention module to capture the space and channel information. • We add focal loss to the overall loss to reduce the impact of class imbalance.

Assessing Macro Disease Index of Wheat Stripe Rust Based on Segformer with Complex Background in the Field.

Wheat stripe rust (WSR) is a foliar disease that causes destructive damage in the wheat production context. Accurately estimating the severity of WSR in the autumn growing stage can help to objectively monitor the disease incidence level of WSR and predict the nationwide disease incidence in the following year, which have great significance for controlling its nationwide spread and ensuring the safety of grain production. In this study, to address the low accuracy and the efficiency of disease index estimation by traditional methods, WSR-diseased areas are segmented based on Segformer, and the macro disease index (MDI) is automatically calculated for the measurement of canopy-scale disease incidence. The results obtained with different semantic segmentation algorithms, loss functions, and data sets are compared for the segmentation effect, in order to address the severe class imbalance in disease region segmentation. We find that: (1) The results of the various models differed significantly, with Segformer being the best algorithm for WSR segmentation (rust class F1 score = 72.60%), based on the original data set; (2) the imbalanced nature of the data has a significant impact on the identification of the minority class (i.e., the rust class), for which solutions based on loss functions and re-weighting of the minority class are ineffective; (3) data augmentation of the minority class or under-sampling of the original data set to increase the proportion of the rust class greatly improved the F1-score of the model (rust class F1 score = 86.6%), revealing that re-sampling is a simple and effective approach to alleviating the class imbalance problem. Finally, the MDI was used to evaluate the models based on the different data sets, where the model based on the augmented data set presented the best performance (R2 = 0.992, RMSE = 0.008). In conclusion, the deep-learning-based semantic segmentation method, and the corresponding optimization measures, applied in this study allow us to achieve pixel-level accurate segmentation of WSR regions on wheat leaves, thus enabling accurate assessment of the degree of WSR disease under complex backgrounds in the field, consequently providing technical support for field surveys and calculation of the disease level.

Active Boundary Loss for Semantic Segmentation

This paper proposes a novel active boundary loss for semantic segmentation. It can progressively encourage the alignment between predicted boundaries and ground-truth boundaries during end-to-end training, which is not explicitly enforced in commonly used cross-entropy loss. Based on the predicted boundaries detected from the segmentation results using current network parameters, we formulate the boundary alignment problem as a differentiable direction vector prediction problem to guide the movement of predicted boundaries in each iteration. Our loss is model-agnostic and can be plugged in to the training of segmentation networks to improve the boundary details. Experimental results show that training with the active boundary loss can effectively improve the boundary F-score and mean Intersection-over-Union on challenging image and video object segmentation datasets.

Uncertainty Estimation via Response Scaling for Pseudo-Mask Noise Mitigation in Weakly-Supervised Semantic Segmentation

Weakly-Supervised Semantic Segmentation (WSSS) segments objects without heavy burden of dense annotation. While as a price, generated pseudo-masks exist obvious noisy pixels, which result in sub-optimal segmentation models trained over these pseudo-masks. But rare studies notice or work on this problem, even these noisy pixels are inevitable after their improvements on pseudo-mask. So we try to improve WSSS in the aspect of noise mitigation. And we observe that many noisy pixels are of high confidences, especially when the response range is too wide or narrow, presenting an uncertain status. Thus, in this paper, we simulate noisy variations of response by scaling the prediction map in multiple times for uncertainty estimation. The uncertainty is then used to weight the segmentation loss to mitigate noisy supervision signals. We call this method URN, abbreviated from Uncertainty estimation via Response scaling for Noise mitigation. Experiments validate the benefits of URN, and our method achieves state-of-the-art results at 71.2% and 41.5% on PASCAL VOC 2012 and MS COCO 2014 respectively, without extra models like saliency detection. Code is available at https://github.com/XMed-Lab/URN.

WaferSegClassNet - A light-weight network for classification and segmentation of semiconductor wafer defects

As the integration density and design intricacy of semiconductor wafers increase, the magnitude and complexity of defects in them are also on the rise. Since the manual inspection of wafer defects is costly, an automated artificial intelligence (AI) based computer-vision approach is highly desired. The previous works on defect analysis have several limitations, such as low accuracy and the need for separate models for classification and segmentation. For analyzing mixed-type defects, some previous works require separately training one model for each defect type, which is non-scalable. In this paper, we present WaferSegClassNet (WSCN), a novel network based on encoder-decoder architecture. WSCN performs simultaneous classification and segmentation of both single and mixed-type wafer defects. WSCN uses a “shared encoder” for classification, and segmentation, which allows training WSCN end-to-end. We use N-pair contrastive loss to first pretrain the encoder and then use BCE-Dice loss for segmentation, and categorical cross-entropy loss for classification. Use of N-pair contrastive loss helps in better embedding representation in the latent dimension of wafer maps. WSCN has a model size of only 0.51MB and performs only 0.2 M FLOPS. Thus, it is much lighter than other state-of-the-art models. Also, it requires only 150 epochs for convergence, compared to 4000 epochs needed by a previous work. We evaluate our model on the MixedWM38 dataset, which has 38,015 images. WSCN achieves an average classification accuracy of 98.2% and a dice coefficient of 0.9999. We are the first to show segmentation results on the MixedWM38 dataset. The source code can be obtained from https://github.com/ckmvigil/WaferSegClassNet.

A Fully Automated Multimodal MRI-Based Multi-Task Learning for Glioma Segmentation and IDH Genotyping.

The accurate prediction of isocitrate dehydrogenase (IDH) mutation and glioma segmentation are important tasks for computer-aided diagnosis using preoperative multimodal magnetic resonance imaging (MRI). The two tasks are ongoing challenges due to the significant inter-tumor and intra-tumor heterogeneity. The existing methods to address them are mostly based on single-task approaches without considering the correlation between the two tasks. In addition, the acquisition of IDH genetic labels is expensive and costly, resulting in a limited number of IDH mutation data for modeling. To comprehensively address these problems, we propose a fully automated multimodal MRI-based multi-task learning framework for simultaneous glioma segmentation and IDH genotyping. Specifically, the task correlation and heterogeneity are tackled with a hybrid CNN-Transformer encoder that consists of a convolutional neural network and a transformer to extract the shared spatial and global information learned from a decoder for glioma segmentation and a multi-scale classifier for IDH genotyping. Then, a multi-task learning loss is designed to balance the two tasks by combining the segmentation and classification loss functions with uncertain weights. Finally, an uncertainty-aware pseudo-label selection is proposed to generate IDH pseudo-labels from larger unlabeled data for improving the accuracy of IDH genotyping by using semi-supervised learning. We evaluate our method on a multi-institutional public dataset. Experimental results show that our proposed multi-task network achieves promising performance and outperforms the single-task learning counterparts and other existing state-of-the-art methods. With the introduction of unlabeled data, the semi-supervised multi-task learning framework further improves the performance of glioma segmentation and IDH genotyping. The source codes of our framework are publicly available at https://github.com/miacsu/MTTU-Net.git.

GapLoss: A Loss Function for Semantic Segmentation of Roads in Remote Sensing Images

At present, road continuity is a major challenge, and it is difficult to extract the centerline vector of roads, especially when the road view is obstructed by trees or other structures. Most of the existing research has focused on optimizing the available deep-learning networks. However, the segmentation accuracy is also affected by the loss function. Currently, little research has been published on road segmentation loss functions. To resolve this problem, an attention loss function named GapLoss that can be combined with any segmentation network was proposed. Firstly, a deep-learning network was used to obtain a binary prediction mask. Secondly, a vector skeleton was extracted from the prediction mask. Thirdly, for each pixel, eight neighboring pixels with the same value of the pixel were calculated. If the value was 1, then the pixel was identified as the endpoint. Fourth, according to the number of endpoints within a buffered range, each pixel in the prediction image was given a corresponding weight. Finally, the weighted average value of the cross-entropy of all the pixels in the batch was used as the final loss function value. We employed four well-known semantic segmentation networks to conduct comparative experiments on three large datasets. The results showed that, compared to other loss functions, the evaluation metrics after using GapLoss were nearly all improved. From the predicted image, the road prediction by GapLoss was more continuous, especially at intersections and when the road was obscured from view, and the road segmentation accuracy was improved.

A Noise-Resilient Framework for Automatic COVID-19 Pneumonia Lesions Segmentation from CT Images

The coronavirus disease pandemic of 2019 (COVID-19) is sweeping the globe. Medical imaging, such as X-ray and computed tomography (CT), is critical in the global fight against COVID-19, and newly developed artificial intelligence (AI) technologies are enhancing the power of imaging tools and assisting medical specialists. We examine the rapid responses to COVID-19 in the medical imaging community (enabled by AI).Although deep learning algorithms have shown promise in a number of areas, they continue to struggle with noisy-labeled images throughout the training phase. Given that the quality of annotation is inextricably linked to a high level of knowledge, the issue is even more pressing in the medical picture arena. It’s still a big difficulty to get rid of the noise from noisy labels for segmentation tasks without adding more annotations. As a noninvasive imaging technique, computed tomography (CT) can detect certain lung symptoms linked with COVID-19. As a result, CT could be a useful tool for early detection and diagnosis of COVID-19. Despite its benefits, CT may have some imaging characteristics in common with COVID-19 and other kinds of pneumonia, making differentiation challenging. Due to its high power of feature extraction, artificial intelligence (AI) leveraging deep learning technology has recently proven remarkable success in the medical imaging arena. Deep learning was used to detect and distinguish between bacterial and viral pneumonia in paediatric chest radiographs.For the segmentation challenge, we present a novel noise-resistant architecture for learning from noisy labels. To better deal with lesions of varied scales and appearances, we present a unique COVID-19 Pneumonia Lesion segmentation network (COPLE-Net), which is a generalisation of Dice loss for segmentation and Mean Absolute Error (MAE) loss for robustness against noise. The noise-resistant Dice loss and COPLENet are combined with an adaptive self-ensembling architecture for training, in which a student model’s Exponential Moving Average (EMA) is employed as a teacher model that is adaptively updated by suppressing the contribution. In the context of learning from noisy labels for COVID-19 pneumonia lesion segmentation, our system with adaptive self-ensembling outperforms a regular training method and outperforms existing noise-robust training approaches.

MVFusFra: A Multi-View Dynamic Fusion Framework for Multimodal Brain Tumor Segmentation.

Medical practitioners generally rely on multimodal brain images, for example based on the information from the axial, coronal, and sagittal views, to inform brain tumor diagnosis. Hence, to further utilize the 3D information embedded in such datasets, this paper proposes a multi-view dynamic fusion framework (hereafter, referred to as MVFusFra) to improve the performance of brain tumor segmentation. The proposed framework consists of three key building blocks. First, a multi-view deep neural network architecture, which represents multi learning networks for segmenting the brain tumor from different views and each deep neural network corresponds to multi-modal brain images from one single view. Second, the dynamic decision fusion method, which is mainly used to fuse segmentation results from multi-views into an integrated method. Then, two different fusion methods (i.e., voting and weighted averaging) are used to evaluate the fusing process. Third, the multi-view fusion loss (comprising segmentation loss, transition loss, and decision loss) is proposed to facilitate the training process of multi-view learning networks, so as to ensure consistency in appearance and space, for both fusing segmentation results and the training of the learning network. We evaluate the performance of MVFusFra on the BRATS 2015 and BRATS 2018 datasets. Findings from the evaluations suggest that fusion results from multi-views achieve better performance than segmentation results from the single view, and also implying effectiveness of the proposed multi-view fusion loss. A comparative summary also shows that MVFusFra achieves better segmentation performance, in terms of efficiency, in comparison to other competing approaches.

Synergistically segmenting choroidal layer and vessel using deep learning for choroid structure analysis

Objective. The choroid is the most vascularized structure in the human eye, whose layer structure and vessel distribution are both critical for the physiology of the retina, and disease pathogenesis of the eye. Although some works have used graph-based methods or convolutional neural networks to separate the choroid layer from the outer-choroid structure, few works focused on further distinguishing the inner-choroid structure, such as the choroid vessel and choroid stroma. Approach. Inspired by the multi-task learning strategy, in this paper, we propose a segmentation pipeline for choroid analysis which can separate the choroid layer from other structures and segment the choroid vessel synergistically. The key component of this pipeline is the proposed choroidal U-shape network (CUNet), which catches both correlation features and specific features between the choroid layer and the choroid vessel. Then pixel-wise classification is generated based on these two types of features to obtain choroid layer segmentation and vessel segmentation. Besides, the training process of CUNet is supervised by a proposed adaptive multi-task segmentation loss which adds a regularization term that is used to balance the performance of the two tasks. Main results. Experiments show the high performance (4% higher dice score) and less computational complexity (18.85 M lower size) of our proposed strategy. Significance. The high performance and generalization on both choroid layer and vessel segmentation indicate the clinical potential of our proposed pipeline.

CHS-Net: A Deep Learning Approach for Hierarchical Segmentation of COVID-19 via CT Images.

The pandemic of novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) also known as COVID-19 has been spreading worldwide, causing rampant loss of lives. Medical imaging such as computed tomography (CT), X-ray, etc., plays a significant role in diagnosing the patients by presenting the visual representation of the functioning of the organs. However, for any radiologist analyzing such scans is a tedious and time-consuming task. The emerging deep learning technologies have displayed its strength in analyzing such scans to aid in the faster diagnosis of the diseases and viruses such as COVID-19. In the present article, an automated deep learning based model, COVID-19 hierarchical segmentation network (CHS-Net) is proposed that functions as a semantic hierarchical segmenter to identify the COVID-19 infected regions from lungs contour via CT medical imaging using two cascaded residual attention inception U-Net (RAIU-Net) models. RAIU-Net comprises of a residual inception U-Net model with spectral spatial and depth attention network (SSD) that is developed with the contraction and expansion phases of depthwise separable convolutions and hybrid pooling (max and spectral pooling) to efficiently encode and decode the semantic and varying resolution information. The CHS-Net is trained with the segmentation loss function that is the defined as the average of binary cross entropy loss and dice loss to penalize false negative and false positive predictions. The approach is compared with the recently proposed approaches and evaluated using the standard metrics like accuracy, precision, specificity, recall, dice coefficient and Jaccard similarity along with the visualized interpretation of the model prediction with GradCam++ and uncertainty maps. With extensive trials, it is observed that the proposed approach outperformed the recently proposed approaches and effectively segments the COVID-19 infected regions in the lungs.

SCL—Segmentation–Classification combined Loss for surface defect detection

Deep learning for surface defect detection has recently received increased attention in relation to quality control of industrial products. In this paper, we present a combined loss function called Segmentation–Classification combined Loss (SCL). It exploits the binary information (normal/anomalous sample) implicit in the segmentation masks to reduce false positives and false negatives in surface defect detection tasks. We evaluate the method on images in three different domains: defects in electrical commutators, defects in a weakly labeled synthetic dataset, and defects of glass bottles.Our combined loss function outperforms the state of the art in the Kolektor (electrical commutators) dataset and significantly improves the classical Dice segmentation loss function in terms of both average precision, false positives, and false negatives. Moreover, the combined loss function introduced in this paper can be applied in a straightforward manner to any loss-based segmentation model. Finally, in this work we introduce and publish a new dataset for the still poorly explored task of surface defect detection of glass bottles.