Annotated Training Research Articles

Comparing field measurements and annotations as training data for UAV-based detection of standing dead trees in coniferous forest

ABSTRACT Deadwood provides a habitat for a large number of species and acts as a significant carbon storage. Thus, deadwood mapping allows planning various actions related to biodiversity conservation and forest management. Airborne laser scanning (ALS) provides an efficient means for monitoring large forested areas and is thus a potential method for deadwood mapping. Machine learning (ML) can be used for automating deadwood detection from ALS data. However, ML requires training data, collecting of which is laborious using field measurements. This study inspected the feasibility of training data manually annotated from aerial images for the mapping of individual standing dead trees. The standing dead trees were mapped from an ALS dataset collected with an unmanned aerial vehicle (UAV). The study was carried out by performing ML-based standing dead tree detection using annotated training data and field-measured training data, and comparing the performances of the models trained on these two datasets. The study found that using annotated training data improves the performance of standing dead tree detection due to its higher availability compared to field-measured data. For large trees (height > 14 m), the precision, recall, Cohen’s kappa score, and Matthews correlation coefficient achieved by the best classifier trained on annotated data were 0.23, 0.48, 0.17, and 0.20, respectively. In comparison, the corresponding metrics for the best classifier trained on field-measured data were 0.14, 0.52, 0.04, and 0.13. Annotated standing dead tree data is not a representative sample of the true standing dead tree population, as small trees and trees without crowns (snags) can often not be identified from aerial images. However, the study found that identifying such trees is challenging even when using field-measured training data, and thus using annotated training data does not bias results. In general, the results of the study showed that ALS-based standing dead tree detection should focus on the detection of large trees.

Can We Replace Real-World With Synthetic Data in Deep Learning-Based ADAS Algorithm Development?

Deep learning models require vast amounts of annotated training data. Gathering and annotating the data from the real world is an expensive and time-consuming process. Thus, synthetically generated data is being researched more and more. This paper tries to answer the question of whether and to what extent synthetically generated data can help in developing the advanced driver-assistance system (ADAS) algorithms for autonomous vehicles, for the object detection task based on deep learning. Two state-of-the-art deep learning object detectors were trained on various combinations of real-world and synthetic data. A total of 12 detectors were tested on real-world test images. Results show that synthetic data can contribute to better detector performance until a certain ratio between real-world and synthetic data is reached.

Learning to detect an animal sound from five examples

Automatic detection and classification of animal sounds has many applications in biodiversity monitoring and animal behavior. In the past twenty years, the volume of digitised wildlife sound available has massively increased, and automatic classification through deep learning now shows strong results. However, bioacoustics is not a single task but a vast range of small-scale tasks (such as individual ID, call type, emotional indication) with wide variety in data characteristics, and most bioacoustic tasks do not come with strongly-labelled training data. The standard paradigm of supervised learning, focussed on a single large-scale dataset and/or a generic pre-trained algorithm, is insufficient. In this work we recast bioacoustic sound event detection within the AI framework of few-shot learning. We adapt this framework to sound event detection, such that a system can be given the annotated start/end times of as few as 5 events, and can then detect events in long-duration audio—even when the sound category was not known at the time of algorithm training. We introduce a collection of open datasets designed to strongly test a system's ability to perform few-shot sound event detections, and we present the results of a public contest to address the task. Our analysis shows that prototypical networks are a very common used strategy and they perform well when enhanced with adaptations for general characteristics of animal sounds. However, systems with high time resolution capabilities perform the best in this challenge. We demonstrate that widely-varying sound event durations are an important factor in performance, as well as non-stationarity, i.e. gradual changes in conditions throughout the duration of a recording. For fine-grained bioacoustic recognition tasks without massive annotated training data, our analysis demonstrate that few-shot sound event detection is a powerful new method, strongly outperforming traditional signal-processing detection methods in the fully automated scenario.

Wav2DDK: Analytical and Clinical Validation of an Automated Diadochokinetic Rate Estimation Algorithm on Remotely Collected Speech.

Oral diadochokinesis is a useful task in assessment of speech motor function in the context of neurological disease. Remote collection of speech tasks provides a convenient alternative to in-clinic visits, but scoring these assessments can be a laborious process for clinicians. This work describes Wav2DDK, an automated algorithm for estimating the diadochokinetic (DDK) rate on remotely collected audio from healthy participants and participants with amyotrophic lateral sclerosis (ALS). Wav2DDK was developed using a corpus of 970 DDK assessments from healthy and ALS speakers where ground truth DDK rates were provided manually by trained annotators. The clinical utility of the algorithm was demonstrated on a corpus of 7,919 assessments collected longitudinally from 26 healthy controls and 82 ALS speakers. Corpora were collected via the participants' own mobile device, and instructions for speech elicitation were provided via a mobile app. DDK rate was estimated by parsing the character transcript from a deep neural network transformer acoustic model trained on healthy and ALS speech. Algorithm estimated DDK rates are highly accurate, achieving .98 correlation with manual annotation, and an average error of only 0.071 syllables per second. The rate exactly matched ground truth for 83% of files and was within 0.5 syllables per second for 95% of files. Estimated rates achieve a high test-retest reliability (r = .95) and show good correlation with the revised ALS functional rating scale speech subscore (r = .67). We demonstrate a system for automated DDK estimation that increases efficiency of calculation beyond manual annotation. Thorough analytical and clinical validation demonstrates that the algorithm is not only highly accurate, but also provides a convenient, clinically relevant metric for tracking longitudinal decline in ALS, serving to promote participation and diversity of participants in clinical research. https://doi.org/10.23641/asha.23787033.

Few-shot multi-view object classification via dual augmentation network

Existing multi-view object classification algorithms usually rely on sufficient labeled multi-view objects, which substantially restricts their scalability to novel classes with few annotated training samples in real-world applications. Aiming to go beyond these limitations, we explore a novel yet challenging task, few-shot multi-view object classification (FS-MVOC), which expects the network to build its classification ability efficiently based on limited labeled multi-view objects. To this end, we design a dual augmentation network (DANet) to provide excellent performance for the under-explored FS-MVOC task. On the one hand, we employ an attention-guided multi-view representation augmentation (AMRA) strategy to help the model focus on salient features and suppress unnecessary ones on multiple views of multi-view objects, resulting in more discriminative multi-view representations. On the other hand, during the meta-training stage, we adopt the category prototype augmentation (CPA) strategy to improve the class-representativeness of each prototype and increase the inter-prototype difference by injecting Gaussian noise in the deep feature space. Extensive experiments on the benchmark datasets (Meta-ModelNet and Meta-ShapeNet) indicate the effectiveness and robustness of DANet.

Zero‐shot temporal event localisation: Label‐free, training‐free, domain‐free

AbstractTemporal event localisation (TEL) has recently attracted increasing attention due to the rapid development of video platforms. Existing methods are based on either fully/weakly supervised or unsupervised learning, and thus they rely on expensive data annotation and time‐consuming training. Moreover, these models, which are trained on specific domain data, limit the model generalisation to data distribution shifts. To cope with these difficulties, the authors propose a zero‐shot TEL method that can operate without training data or annotations. Leveraging large‐scale vision and language pre‐trained models, for example, CLIP, we solve the two key problems: (1) how to find the relevant region where the event is likely to occur; (2) how to determine event duration after we find the relevant region. Query guided optimisation for local frame relevance relying on the query‐to‐frame relationship is proposed to find the most relevant frame region where the event is most likely to occur. Proposal generation method relying on the frame‐to‐frame relationship is proposed to determine the event duration. The authors also propose a greedy event sampling strategy to predict multiple durations with high reliability for the given event. The authors’ methodology is unique, offering a label‐free, training‐free, and domain‐free approach. It enables the application of TEL purely at the testing stage. The practical results show it achieves competitive performance on the standard Charades‐STA and ActivityCaptions datasets.

Deep unsupervised learning for 3D ALS point clouds change detection

Change detection from traditional 2D optical images has limited capability to model the changes in the height or shape of objects. Change detection using 3D point cloud from photogrammetry or LiDAR surveying can fill this gap by providing critical depth information. While most existing machine learning based 3D point cloud change detection methods are supervised, they severely depend on the availability of annotated training data, which is in practice a critical point. To circumnavigate this dependence, we propose an unsupervised 3D point cloud change detection method mainly based on self-supervised learning using deep clustering and contrastive learning. The proposed method also relies on an adaptation of deep change vector analysis to 3D point cloud via nearest point comparison. Experiments conducted on an aerial LiDAR survey dataset show that the proposed method obtains higher performance in comparison to the traditional unsupervised methods, with a gain of about 9% in mean accuracy (to reach more than 85%). Thus, it appears to be a relevant choice in scenario where prior knowledge (labels) is not ensured. The code will be made available at https://github.com/IdeGelis/torch-points3d-SSL-DCVA.

Weakly Supervised MRI Slice-Level Deep Learning Classification of Prostate Cancer Approximates Full Voxel- and Slice-Level Annotation: Effect of Increasing Training Set Size.

Weakly supervised learning promises reduced annotation effort while maintaining performance. To compare weakly supervised training with full slice-wise annotated training of a deep convolutional classification network (CNN) for prostate cancer (PC). Retrospective. One thousand four hundred eighty-nine consecutive institutional prostate MRI examinations from men with suspicion for PC (65 ± 8 years) between January 2015 and November 2020 were split into training (N = 794, enriched with 204 PROSTATEx examinations) and test set (N = 695). 1.5 and 3T, T2-weighted turbo-spin-echo and diffusion-weighted echo-planar imaging. Histopathological ground truth was provided by targeted and extended systematic biopsy. Reference training was performed using slice-level annotation (SLA) and compared to iterative training utilizing patient-level annotations (PLAs) with supervised feedback of CNN estimates into the next training iteration at three incremental training set sizes (N = 200, 500, 998). Model performance was assessed by comparing specificity at fixed sensitivity of 0.97 [254/262] emulating PI-RADS ≥ 3, and 0.88-0.90 [231-236/262] emulating PI-RADS ≥ 4 decisions. Receiver operating characteristic (ROC) and area under the curve (AUC) was compared using DeLong and Obuchowski test. Sensitivity and specificity were compared using McNemar test. Statistical significance threshold was P = 0.05. Test set (N = 695) ROC-AUC performance of SLA (trained with 200/500/998 exams) was 0.75/0.80/0.83, respectively. PLA achieved lower ROC-AUC of 0.64/0.72/0.78. Both increased performance significantly with increasing training set size. ROC-AUC for SLA at 500 exams was comparable to PLA at 998 exams (P = 0.28). ROC-AUC was significantly different between SLA and PLA at same training set sizes, however the ROC-AUC difference decreased significantly from 200 to 998 training exams. Emulating PI-RADS ≥ 3 decisions, difference between PLA specificity of 0.12 [51/433] and SLA specificity of 0.13 [55/433] became undetectable (P = 1.0) at 998 exams. Emulating PI-RADS ≥ 4 decisions, at 998 exams, SLA specificity of 0.51 [221/433] remained higher than PLA specificity at 0.39 [170/433]. However, PLA specificity at 998 exams became comparable to SLA specificity of 0.37 [159/433] at 200 exams (P = 0.70). Weakly supervised training of a classification CNN using patient-level-only annotation had lower performance compared to training with slice-wise annotations, but improved significantly faster with additional training data. 3 TECHNICAL EFFICACY: Stage 2.

MTMVC: Semi-supervised 3D hand pose estimation using multi-task and multi-view consistency

The high performance of state-of-the-art deep learning methods for 3D hand pose estimation heavily depends on a large annotated training set. However, it is difficult and time-consuming to obtain the annotations for 3D hand poses. To leverage unannotated images to reduce the annotation cost, we propose a semi-supervised method based on Multi-Task and Multi-View Consistency (MTMVC) for hand pose estimation. First, we obtain the joints based on heatmap prediction and coordinate regression parallelly and encourage their consistency. Second, we introduce multi-view consistency to encourage the predicted poses to be rotation-invariant. Thirdly, to make the network pay more attention to the hand region, we propose a spatially weighted consistency. Experiments on four public datasets showed that our proposed MTMVC outperformed existing semi-supervised hand pose estimation methods, and by only using half of the annotations, the accuracy of our method was comparable to those of several state-of-the-art fully supervised methods.

In vitro laser scanning confocal microscopy and unsupervised segmentation: Quantification of cytosolic calcium and RNA distribution in hypoxic neurons.

The mimicry of neurodegenerative diseases in vitro can be observed through the induction of chronic hypoxia, and the impact of this stress is monitored using multiplexed imaging techniques. While laser scanning confocal microscopy (LSCM) is a valuable tool for observing single neurons under degenerative conditions, accurately quantifying RNA distribution and cell size by deep learning tools remains challenging due to the lack of annotated training datasets. To address this, we propose a framework that combines 3D tracking of RNA distribution and cell size identification using unsupervised image segmentation. Additionally, we quantified the calcium level in neurons using fluorescent microscopy using unsupervised image segmentation. First, we performed imaging of neuronal morphology using differential interference contrast (DIC) optics and RNA/calcium level imaging using fluorescent microscopy. Next, we performed k-means clustering-based cell segmentation. The results show that our framework can distinguish between distinct neuronal states under control and chronic hypoxic conditions. The analysis reveals that hypoxia induces a significant increase in cytosolic calcium level, reduction in neuron diameter, and alterations in RNA distribution.Clinical Relevance- The proposed framework is crucial to study the neurodegeneration process and evaluating the efficacy of neuroprotective drugs through image analysis.

Classification and detection of diabetic retinopathy based on multi-scale shallow neural network

<p class="15">The high-quality annotated training samples in medical image processing have limited the development of deep neural networks in their field. This paper designs and proposes an integrated method for classifying and detecting diabetic retinopathy based on a multi-scale shallow neural network. The method consists of multiple shallow neural network base learners, which extract pathological features under different receptive fields. The integrated learning strategy proposed is used to optimize the integration and finally realize the classification and detection of diabetic retinopathy. In addition, to verify the effectiveness of the method in this paper on a small sample data-set, based on the two-dimensional entropy of the image, multiple sub-datasets are constructed for verification. The results show that, compared with the existing methods, the integrated method for the classification and detection of diabetic retinopathy proposed in this paper has a good detection effect on a small sample data-set.</p>

RegEMR: a natural language processing system to automatically identify premature ovarian decline from Chinese electronic medical records

BackgroundThe ovarian reserve is a reservoir for reproductive potential. In clinical practice, early detection and treatment of premature ovarian decline characterized by abnormal ovarian reserve tests is regarded as a critical measure to prevent infertility. However, the relevant data are typically stored in an unstructured format in a hospital’s electronic medical record (EMR) system, and their retrieval requires tedious manual abstraction by domain experts. Computational tools are therefore needed to reduce the workload.MethodsWe presented RegEMR, an artificial intelligence tool composed of a rule-based natural language processing (NLP) extractor and a knowledge-based disease scoring model, to automatize the screening procedure of premature ovarian decline using Chinese reproductive EMRs. We used regular expressions (REs) as a text mining method and explored whether REs automatically synthesized by the genetic programming-based online platform RegexGenerator + + could be as effective as manually formulated REs. We also investigated how the representativeness of the learning corpus affected the performance of machine-generated REs. Additionally, we translated the clinical diagnostic criteria into a programmable disease diagnostic model for disease scoring and risk stratification. Four hundred outpatient medical records were collected from a Chinese fertility center. Manual review served as the gold standard, and fivefold cross-validation was used for evaluation.ResultsThe overall F-score of manually built REs was 0.9444 (95% CI 0.9373 to 0.9515), with no significant difference (paired t test p > 0.05) compared with machine-generated REs that could be affected by training set sizes and annotation portions. The extractor performed effectively in automatically tracing the dynamic changes in hormone levels (F-score 0.9518–0.9884) and ultrasonographic measures (F-score 0.9472–0.9822). Applying the extracted information to the proposed diagnostic model, the program obtained an accuracy of 0.98 and a sensitivity of 0.93 in risk screening. For each specific disease, the automatic diagnosis in 76% of patients was consistent with that of the clinical diagnosis, and the kappa coefficient was 0.63.ConclusionA Chinese NLP system named RegEMR was developed to automatically identify high risk of early ovarian aging and diagnose related diseases from Chinese reproductive EMRs. We hope that this system can aid EMR-based data collection and clinical decision support in fertility centers.

Transferring from Textual Entailment to Biomedical Named Entity Recognition.

Biomedical Named Entity Recognition (BioNER) aims at identifying biomedical entities such as genes, proteins, diseases, and chemical compounds in the given textual data. However, due to the issues of ethics, privacy, and high specialization of biomedical data, BioNER suffers from the more severe problem of lacking in quality labeled data than the general domain especially for the token-level. Facing the extremely limited labeled biomedical data, this work studies the problem of gazetteer-based BioNER, which aims at building a BioNER system from scratch. It needs to identify the entities in the given sentences when we have zero token-level annotations for training. Previous works usually use sequential labeling models to solve the NER or BioNER task and obtain weakly labeled data from gazetteers when we don't have full annotations. However, these labeled data are quite noisy since we need the labels for each token and the entity coverage of the gazetteers is limited. Here we propose to formulate the BioNER task as a Textual Entailment problem and solve the task via Textual Entailment with Dynamic Contrastive learning (TEDC). TEDC not only alleviates the noisy labeling issue, but also transfers the knowledge from pre-trained textual entailment models. Additionally, the dynamic contrastive learning framework contrasts the entities and non-entities in the same sentence and improves the model's discrimination ability. Experiments on two real-world biomedical datasets show that TEDC can achieve state-of-the-art performance for gazetteer-based BioNER.

Multi-path decoder U-Net: A weakly trained real-time segmentation network for object detection and localization in ultrasound scans.

Detecting and localizing an anatomical structure of interest within the field of view of an ultrasound scan is an essential step in many diagnostic and therapeutic procedures. However, ultrasound scans suffer from high levels of variabilities across sonographers and patients, making it challenging for sonographers to accurately identify and locate these structures without extensive experience. Segmentation-based convolutional neural networks (CNNs) have been proposed as a solution to assist sonographers in this task. Despite their accuracy, these networks require pixel-wise annotations for training; an expensive and labor-intensive operation that requires the expertise of an experienced practitioner to identify the precise outline of the structures of interest. This complicates, delays, and increases the cost of network training and deployment. To address this problem, we propose a multi-path decoder U-Net architecture that is trained on bounding box segmentation maps; not requiring pixel-wise annotations. We show that the network can be trained on small training sets, which is the case in medical imaging datasets; reducing the cost and time needed for deployment and use in clinical settings. The multi-path decoder design allows for better training of deeper layers and earlier attention to the target anatomical structures of interest. This architecture offers up to a 7% relative improvement compared to the U-Net architecture in localization and detection performance, with an increase of only 0.75% in the number of parameters. Its performance is on par with, or slightly better than, the more computationally expensive U-Net++, which has 20% more parameters; making the proposed architecture a more computationally efficient alternative for real-time object detection and localization in ultrasound scans.

DeepInfusion: A dynamic infusion based-neuro-symbolic AI model for segmentation of intracranial aneurysms

The detection and segmentation of cerebral aneurysms is a crucial step in the development of a clinical decision support system for estimating aneurysm rupture risk. However, accurately identifying and segmenting regions of interest in two-dimensional (2D) medical images is often challenging, particularly when using deep learning (DL) methods on small datasets with limited annotated data. The accuracy of DL approaches is often affected by the availability of large, annotated training datasets that are required for effective deep learning. Additionally, when using DL to differentiate aneurysms from arterial loops in 2D DSA images, DL can fail to detect aneurysms in areas where dye concentration is low. To address these issues and enhance the reliability and accuracy of aneurysm detection and segmentation methods, incorporating medical expert-advised, hand-crafted features can provide a clinical perspective to DL methods. This approach can help to improve the performance of DL methods by providing additional information that is not captured in the data. To this end, a novel Neuro-symbolic AI-based DeepInfusion model is proposed which allows for the infusion of human intellect through hand-crafted features into deep neural networks (DNNs), thus combining the strengths of DL with the knowledge and expertise of medical professionals. The proposed approach includes a novel technique for dynamic layer selection and feature weight adjustment during the model infusion process. The performance of the DeepInfusion model is evaluated on an in-house prepared dataset of 409 DSA images, and experimental results demonstrate the effectiveness of the proposed method for the segmentation of cerebral aneurysms. The model achieves an IOU score of 96.76% and an F1-score of 94.15% on unseen DSA images. The model is also tested on two publicly available datasets of Kvasir-SEG polyp and DRIVE for vessel segmentation of retinal images. The results show a significant improvement compared to existing methods, which indicates the generalizability of the approach in medical segmentation. The complete code for DeepInfusion is available on our GitHub repository at https://github.com/smileslab/deep-infusion/blob/main/deepinfusion.ipynb.

LWSIS: LiDAR-Guided Weakly Supervised Instance Segmentation for Autonomous Driving

Image instance segmentation is a fundamental research topic in autonomous driving, which is crucial for scene understanding and road safety. Advanced learning-based approaches often rely on the costly 2D mask annotations for training. In this paper, we present a more artful framework, LiDAR-guided Weakly Supervised Instance Segmentation (LWSIS), which leverages the off-the-shelf 3D data, i.e., Point Cloud, together with the 3D boxes, as natural weak supervisions for training the 2D image instance segmentation models. Our LWSIS not only exploits the complementary information in multimodal data during training but also significantly reduces the annotation cost of the dense 2D masks. In detail, LWSIS consists of two crucial modules, Point Label Assignment (PLA) and Graph-based Consistency Regularization (GCR). The former module aims to automatically assign the 3D point cloud as 2D point-wise labels, while the atter further refines the predictions by enforcing geometry and appearance consistency of the multimodal data. Moreover, we conduct a secondary instance segmentation annotation on the nuScenes, named nuInsSeg, to encourage further research on multimodal perception tasks. Extensive experiments on the nuInsSeg, as well as the large-scale Waymo, show that LWSIS can substantially improve existing weakly supervised segmentation models by only involving 3D data during training. Additionally, LWSIS can also be incorporated into 3D object detectors like PointPainting to boost the 3D detection performance for free. The code and dataset are available at https://github.com/Serenos/LWSIS.

AUC Maximization for Low-Resource Named Entity Recognition

Current work in named entity recognition (NER) uses either cross entropy (CE) or conditional random fields (CRF) as the objective/loss functions to optimize the underlying NER model. Both of these traditional objective functions for the NER problem generally produce adequate performance when the data distribution is balanced and there are sufficient annotated training examples. But since NER is inherently an imbalanced tagging problem, the model performance under the low-resource settings could suffer using these standard objective functions. Based on recent advances in area under the ROC curve (AUC) maximization, we propose to optimize the NER model by maximizing the AUC score. We give evidence that by simply combining two binary-classifiers that maximize the AUC score, significant performance improvement over traditional loss functions is achieved under low-resource NER settings. We also conduct extensive experiments to demonstrate the advantages of our method under the low-resource and highly-imbalanced data distribution settings. To the best of our knowledge, this is the first work that brings AUC maximization to the NER setting. Furthermore, we show that our method is agnostic to different types of NER embeddings, models and domains. The code of this work is available at https://github.com/dngu0061/NER-AUC-2T.

A Graph Fusion Approach for Cross-Lingual Machine Reading Comprehension

Although great progress has been made for Machine Reading Comprehension (MRC) in English, scaling out to a large number of languages remains a huge challenge due to the lack of large amounts of annotated training data in non-English languages. To address this challenge, some recent efforts of cross-lingual MRC employ machine translation to transfer knowledge from English to other languages, through either explicit alignment or implicit attention. For effective knowledge transition, it is beneficial to leverage both semantic and syntactic information. However, the existing methods fail to explicitly incorporate syntax information in model learning. Consequently, the models are not robust to errors in alignment and noises in attention. In this work, we propose a novel approach, which jointly models the cross-lingual alignment information and the mono-lingual syntax information using a graph. We develop a series of algorithms, including graph construction, learning, and pre-training. The experiments on two benchmark datasets for cross-lingual MRC show that our approach outperforms all strong baselines, which verifies the effectiveness of syntax information for cross-lingual MRC.

Adapting deep learning models to new acoustic environments - A case study on the North Atlantic right whale upcall

Passive acoustic monitoring is increasingly being used for studying marine mammals, leading to the accumulation of large acoustic datasets. Analyzing these datasets becomes impractical without automated detection and classification software. Detectors and classifiers based on deep neural networks have shown great potential, but their performance is often limited by the availability of sufficient quantities of annotated training samples, and their application restricted to the specific acoustic environment(s) from which their training data were collected. We address these limitations by employing transfer learning, a deep learning concept whereby knowledge from a source domain is transferred to a target domain. Specifically, we considered two different underwater acoustic environments as the source and target domains. The objective was to use a deep neural network that had been trained in one environment with abundant annotated training samples, and optimize its performance in the other environment where the annotated training samples were limited. Training and testing were conducted using three acoustic datasets containing North Atlantic right whale (Eubalaena glacialis) upcalls. Experiments show that adapting a trained model to the new environment led to a substantial improvement in recall from 70% to 85%, while maintaining a low false-positive rate of less than 5 per hour. The methodology is implemented as an open-source Python tool to facilitate the creation of more tailored deep learning-based acoustic detectors and classifiers for North Atlantic right whale vocalizations and other stereotyped marine mammal calls.

Recent Few-shot Object Detection Algorithms: A Survey with Performance Comparison

The generic object detection (GOD) task has been successfully tackled by recent deep neural networks, trained by an avalanche of annotated training samples from some common classes. However, it is still non-trivial to generalize these object detectors to the novel long-tailed object classes, which have only few labeled training samples. To this end, the Few-Shot Object Detection (FSOD) has been topical recently, as it mimics the humans’ ability of learning to learn and intelligently transfers the learned generic object knowledge from the common heavy-tailed to the novel long-tailed object classes. Especially, the research in this emerging field has been flourishing in recent years with various benchmarks, backbones, and methodologies proposed. To review these FSOD works, there are several insightful FSOD survey articles [ 58 , 59 , 74 , 78 ] that systematically study and compare them as the groups of fine-tuning/transfer learning and meta-learning methods. In contrast, we review the existing FSOD algorithms from a new perspective under a new taxonomy based on their contributions, i.e., data-oriented, model-oriented, and algorithm-oriented. Thus, a comprehensive survey with performance comparison is conducted on recent achievements of FSOD. Furthermore, we also analyze the technical challenges, the merits and demerits of these methods, and envision the future directions of FSOD. Specifically, we give an overview of FSOD, including the problem definition, common datasets, and evaluation protocols. The taxonomy is then proposed that groups FSOD methods into three types. Following this taxonomy, we provide a systematic review of the advances in FSOD. Finally, further discussions on performance, challenges, and future directions are presented.