Computer vision species identification of lichens and bryophytes from biocrusts in Australian drylands

Plants meet machines: Prospects in machine learning for plant biology

Classification of imbalanced multi-class image datasets is a challenging problem in computer vision. Most of the real-world datasets are imbalanced in nature because of the uneven distribution of the samples in each class. The problem with an imbalanced dataset is that the minority class having a smaller number of instance samples is left undetected. Most of the traditional machine learning algorithms can detect the majority class efficiently but lag behind in the efficient detection of the minority class, which ultimately degrades the overall performance of the classification model. In this paper, we have proposed a novel combination of visual codebook generation using deep features with the non-linear Chi2 SVM classifier to tackle the imbalance problem that arises while dealing with multi-class image datasets. The low-level deep features are first extracted by transfer learning using the ResNet-50 pre-trained network, and clustered using k-means. The center of each cluster is a visual word in the codebook. Each image is then translated into a set of features called the Bag-of-Visual-Words (BOVW) derived from the histogram of visual words in the vocabulary. The non-linear Chi2 SVM classifier is found most optimal for classifying the ensuing features, as proved by a detailed empirical analysis. Hence with the right combination of learning tools, we are able to tackle classification of multi-class imbalanced image datasets in an effective manner. This is proved from the higher scores of accuracy, F1-score and AUC metrics in our experiments on two challenging multi-class datasets: Graz-02 and TF-Flowers, as compared to the state-of-the-art methods.

Imbalance Learning and Its Application on Medical Datasets

Artificial Intelligence-related Literature in Transplantation: A Practical Guide.

In the field of supervised learning, the problem of class imbalance is one of the most difficult problems, and has attracted a great deal of research attention in recent years. In an imbalanced dataset, minority classes are those that contain very small numbers of data samples, while the remaining classes have a very large number of data samples. This type of imbalance reduces the predictive performance of machine learning models. There are currently three approaches for dealing with the class imbalance problem: algorithm-level, data-level, and ensemble-based approaches. Of these, data-level approaches are the most widely used, and consist of three sub-categories: under-sampling, oversampling, and hybrid techniques. Oversampling techniques generate synthetic samples for the minority class to balance an imbalanced dataset. However, existing oversampling approaches do not have a strategy for handling noise samples in imbalanced and noisy datasets, which leads to a reduction in the predictive performance of machine learning models. This study therefore proposes a noise-adaptive synthetic oversampling technique (NASOTECH) to deal with the class imbalance problem in imbalanced and noisy datasets. The noise-adaptive synthetic oversampling (NASO) strategy is first introduced, which is used to identify the number of samples generated for each sample in the minority class, based on the concept of the noise ratio. Next, the NASOTECH algorithm is proposed, based on the NASO strategy, to handle the class imbalance problem in imbalanced and noisy datasets. Finally, empirical experiments are conducted on several synthetic and real datasets to verify the effectiveness of the proposed approach. The experimental results confirm that NASOTECH outperforms three state-of-the-art oversampling techniques in terms of accuracy and geometric mean (G-mean) on imbalanced and noisy datasets.

Manual detection of abnormalities accurately in the clinical images like MRIs by the operators is tedious work that requires good experience and knowledge, specifically manually segmenting the brain tumor in the MRI for further diagnosis by the doctor. So, multiple automatic and semi-automatic approaches were developed to automate the process of segmenting the malignant area of the tumor. The major problem which arises to train the model for automatic segmentation of clinical images is the imbalanced data set. An imbalanced clinical data set means the healthy tissues are always far greater than the cancerous tissues. This difference between the majority and minority data in the data sets reduces or adversely affects the accuracy of predicting model due to biased training data sets. So, it becomes a major concern for the various researchers to balance the data before using it to train a particular prediction model, and various data-level and algorithm-levelbased approaches were developed to balance the imbalance data for improving the accuracy of the trained model. In this chapter, the concept and problem of imbalanced data are discussed and various approaches for balancing the data are also highlighted in which one of the state-of-the-art method bagging is discussed in detail.

Clinical abbreviations pose a challenge for clinical decision support systems due to their ambiguity. Additionally, clinical datasets often suffer from class imbalance, hindering the classification of such data. This imbalance leads to classifiers with low accuracy and high error rates. Traditional feature-engineered models struggle with this task, and class imbalance is a known factor that reduces the performance of neural network techniques. This study proposes an attention-based bidirectional long short-term memory (Bi-LSTM) model to improve clinical abbreviation disambiguation in clinical documents. We aim to address the challenges of limited training data and class imbalance by employing data generation techniques like reverse substitution and data augmentation with synonym substitution. We utilise a Bi-LSTM classification model with an attention mechanism to disambiguate each abbreviation. The model's performance is evaluated based on accuracy for each abbreviation. To address the limitations of imbalanced data, we employ data generation techniques to create a more balanced dataset. The evaluation results demonstrate that our data balancing technique significantly improves the model's accuracy by 2.08%. Furthermore, the proposed attention-based Bi-LSTM model achieves an accuracy of 96.09% on the UMN dataset, outperforming state-of-the-art results. Deep neural network methods, particularly Bi-LSTM, offer promising alternatives to traditional feature-engineered models for clinical abbreviation disambiguation. By employing data generation techniques, we can address the challenges posed by limited-resource and imbalanced clinical datasets. This approach leads to a significant improvement in model accuracy for clinical abbreviation disambiguation tasks.

Premise Equisetum is a distinctive vascular plant genus with 15 extant species worldwide. Species identification is complicated by morphological plasticity and frequent hybridization events, leading to a disproportionately high number of misidentified specimens. These may be correctly identified by applying appropriate computer vision tools.MethodsWe hypothesize that aerial stem nodes can provide enough information to distinguish among Equisetum hyemale, E. laevigatum, and E . ×ferrissii, the latter being a hybrid between the other two. An object detector was trained to find nodes on a given image and to distinguish E. hyemale nodes from those of E. laevigatum. A classifier then took statistics from the detection results and classified the given image into one of the three taxa. Both detector and classifier were trained and tested on expert manually annotated images.ResultsIn our exploratory test set of 30 images, our detector/classifier combination identified all 10 E. laevigatum images correctly, as well as nine out of 10 E. hyemale images, and eight out of 10 E. ×ferrissii images, for a 90% classification accuracy.DiscussionOur results support the notion that computer vision may help with the identification of herbarium specimens once enough manual annotations become available.

Aiming to address the problem that the existing methods for detecting sow backfat thickness are stressful, costly, and cannot detect in real time, this paper proposes a non-contact detection method for sow backfat with a residual network based on image segmentation using the feature visualization of neural networks. In this paper, removing the irrelevant information of the image to improve the accuracy of the sow backfat thickness detection model is proposed. The irrelevant features in the corresponding image of the feature map are found to have the same high brightness as the relevant feature regions using feature visualization. An image segmentation algorithm is then used to separate the relevant feature image regions, and the model performance before and after image segmentation is compared to verify the feasibility of this method. In order to verify the generalization ability of the model, five datasets were randomly divided, and the test results show that the coefficients of determination (R2) of the five groups were above 0.89, with a mean value of 0.91, and the mean absolute error (MAE) values were below 0.66 mm, with a mean value of 0.54 mm, indicating that the model has high detection accuracy and strong robustness. In order to explain the high accuracy of the backfat thickness detection model and to increase the credibility of the application of the detection model, using feature visualization, the irrelevant features and related features of the sow back images extracted by the residual network were statistically analyzed, which were the characteristics of the hip edge, the area near the body height point, the area near the backfat thickness measurement point (P2), and the lateral contour edge. The first three points align with the previous research on sow backfat, thus explaining the phenomenon of the high accuracy of the detection model. At the same time, the side contour edge features were found to be effective for predicting the thickness of the back. In order to explore the influence of irrelevant features on the accuracy of the model, UNet was used to segment the image area corresponding to the irrelevant features and obtain the sow contour image, which was used to construct a dorsal fat thickness detection model. The R2 results of the model were above 0.91, with a mean value of 0.94, and the MAE was below 0.65 mm, with a mean value of 0.44 mm. Compared to the test results of the model before segmentation, the average R2 of the model after segmentation increased by 3.3%, and the average MAE decreased by 18.5%, indicating that irrelevant features will reduce the detection accuracy of the model, which can provide a reference for farmers to dynamically monitor the backfat of sows and accurately manage their farms.

A comparative study on noise filtering of imbalanced data sets

We compare species richness of bryophytes and vascular plants in Estonian moist forests and mires. The material was collected from two wetland nature reserves. Bryophyte and vascular plant spe- cies were recorded in 338 homogeneous stands of approximately 1 ha in nine forest and two mire types. Regional species pools for bryophytes and vascular plants were significantly correlated. The correlations between the species richnesses of bryophytes and vascular plants per stand were positive in all community types. The relative richnesses (local richness divided by the regional species pool size) were similar for bryophyte species and for vascular plant species. This shows that on larger scales, conservation of the communities rich in species of one taxonomic plant group, maintains also the species richness of the other. The minimum number of stands needed for the maintenance of the regional species pool of typical species for the every community type was calculated using the species richness accumulation curves. Less stands are needed to maintain the bryophyte species pools (300-5300 for bryophytes and 400-35 000 for vascular plants).

Image classification and segmentation are important tasks in computer vision. ResNet and U-Net are representative networks for image classification and image segmentation, respectively. Although many scholars used to fuse these two networks, most integration focuses on image segmentation with U-Net, overlooking the capabilities of ResNet for image classification. In this paper, we propose a novel U-ResNet structure by combining U-Net’s convolution–deconvolution structure (UBlock) with ResNet’s residual structure (ResBlock) in a parallel manner. This novel parallel structure achieves rapid convergence and high accuracy in image classification and segmentation while also efficiently alleviating the vanishing gradient problem. Specifically, in the UBlock, the pixel-level features of both high- and low-resolution images are extracted and processed. In the ResBlock, a Selected Upsampling (SU) module was introduced to enhance performance on low-resolution datasets, and an improved Efficient Upsampling Convolutional Block (EUCB*) with a Channel Shuffle mechanism was added before the output of the ResBlock to enhance network convergence. Features from both the ResBlock and UBlock were merged for better decision making. This architecture outperformed the state-of-the-art (SOTA) models in both image classification and segmentation tasks on open-source and private datasets. Functions of individual modules were further verified via ablation studies. The superiority of the proposed U-ResNet displays strong feasibility and potential for advanced cross-paradigm tasks in computer vision.

In grassland communities vascular plants and bryophytes form two distinct layers. In order to understand the factors responsible for plant community structure, more information about interactions between these plant groups is needed. Often negative correlations between vascular plant and bryophyte covers have been reported, suggesting competition. Here we tested experimentally whether different grassland vascular plant species (Trifolium pratense, Festuca pratensis, Prunella vulgaris) had different influences on the cover of two bryophyte species (Rhytidiadelphus squarrosus, Brachythecium rutabulum). In a two-year garden pot experiment one bryophyte species and one vascular plant species were planted per pot. Bryophytes were planted at a constant density, vascular plants in four densities. The cover of both bryophyte species increased with increasing vascular plant cover, showing the facilitative effect of vascular plants through creating better microclimate, e.g., optimising temperature. Bryophyte responses to vascular plant species were species-specific. Festuca had significantly positive effects on both bryophyte species in the second year, and Trifolium on Brachythecium in both years, whereas Prunella had no significant effect on bryophytes. The facilitative effect of vascular plants was stronger at the second experimental year. In summary, the biotic effects between bryophytes and grassland vascular plants are species-specific and positive interactions are prevailing at low vascular plant densities.

Image segmentation is a crucial task in computer vision and image processing. It is widely used in many necessary fields, such as scene understanding, medical image analysis, robot perception, video surveillance, augmented reality, and image compression. Numerous algorithms for image segmentation have been proposed, demonstrating their unique advantages and limitations in their respective application scenarios. In image processing and pattern recognition, the importance and criticality of image segmentation are self-evident. Its core task is to divide the entire image into several regions with specific meaning and define a category label for each area. In recent years, convolutional neural networks (CNN) have performed well in image segmentation and have become one of the most popular and widely used models. This paper focuses on changing the model scale, which significantly impacts the segmentation results by changing the size of the data set used to train the model. This paper aims to explore the impact of data volume on model performance. For example, will the segmentation results become more accurate as the model scale increases? This paper first created and trained a CNN model using different scales. In each training, this paper trains the model for 50 epochs, which can significantly improve the reliability and accuracy of the experimental results. Next, this paper segments the test image, analyzes the segmentation effect, and further explores the relationship between parameters scale and model performance. This research will provide new ideas and references for optimizing image segmentation.

Occupancy detection for large buildings enables optimised control of indoor systems based on occupant presence, reducing the energy costs of heating and cooling. Through machine learning models, occupancy detection is achieved with an accuracy of over 95%. However, to achieve this, large amounts of data are collected with little consideration of which of the collected data are most useful to the task. This paper demonstrates methods to identify if data may be removed from the imbalanced time-series training datasets to optimise the training process and model performance. It also describes how the calculation of the class density of a dataset may be used to identify if a dataset is applicable for data reduction, and how dataset fusion may be used to combine occupancy datasets. The results show that over 50% of a training dataset may be removed from imbalanced datasets while maintaining performance, reducing training time and energy cost by over 40%. This indicates that a data-centric approach to developing artificial intelligence applications is as important as selecting the best model.