HARNet in deep learning approach-a systematic survey.

A comprehensive examination of human action recognition (HAR) methodologies situated at the convergence of deep learning and computer vision is the subject of this article. We examine the progression from handcrafted feature-based approaches to end-to-end learning, with a particular focus on the significance of large-scale datasets. By classifying research paradigms, such as temporal modelling and spatial features, our proposed taxonomy illuminates the merits and drawbacks of each. We specifically present HARNet, an architecture for Multi-Model Deep Learning that integrates recurrent and convolutional neural networks while utilizing attention mechanisms to improve accuracy and robustness. The VideoMAE v2 method ( https://github.com/OpenGVLab/VideoMAEv2 ) has been utilized as a case study to illustrate practical implementations and obstacles. For researchers and practitioners interested in gaining a comprehensive understanding of the most recent advancements in HAR as they relate to computer vision and deep learning, this survey is an invaluable resource.

Pronounced Self-Induced Diastereomeric Anisochronism in Anisidine Amino Acid Diamides.

The emergent properties resulting from selective supramolecular interactions are of significant importance for materials and chemical systems. For the directed use of such properties, a fundamental understanding of the interaction mechanism and the resulting mode of function is necessary for a tailored design. The self-induced diastereomeric anisochronism effect (SIDA), which occurs in the intermolecular interaction of chiral molecules, generates unique properties such as chiral self-recognition and nonlinear effects. Here we show that anisidine amino acid diamides lead to extraordinary signal splitting in NMR spectra through supramolecular interaction and homochiral self-recognition. By systematic experiments we have investigated the underlying SIDA effect, explored its limits and finally successfully utilized it in the determination of enantiomeric ratios by NMR spectroscopy of chiral 'SIDA-inactive' compounds such as thalidomide.

Co-assembled Nanoparticles Towards Multi-target Combinational Therapy of Alzheimer's Disease by Making Full Use of Molecular Recognition and Self-assembly.

The complex pathologies in Alzheimer's disease (AD) severely limits the effectiveness of single-target pharmic interventions, thus necessitating multi-pronged therapeutic strategies. While flexibility is essentially demanded in constructing such multi-target systems, for achieving optimal synergies and also accommodating the inherent heterogeneity within AD. Utilizing the dynamic reversibility of supramolecular strategy for conferring sufficient tunability in component substitution and proportion adjustment, amphiphilic calixarenes are poised to be a privileged molecular tool for facilely achieving function integration. Herein, taking ß-amyloid (Aß) fibrillation and oxidative stress as model combination pattern, we proposed a supramolecular multifunctional integration by co-assembling guanidinium-modified calixarene with ascorbyl palmitate and loading dipotassium phytate within calixarene cavity. Serial pivotal events can be simultaneously addressed by this versatile system, including (1) inhibition of Aß production and aggregation, (2) disintegration of Aß fibrils, (3) acceleration of Aß metabolic clearance, and (4) regulation of oxidative stress, which is verified to significantly ameliorate the cognitive impairment of 5×FAD mice, with reduced Aß plaque content, neuroinflammation and neuronal apoptosis. Confronted with the extremely intricate clinical realities of AD, the strategy presented here exhibits ample adaptability for necessary alterations on combinations, thereby may immensely expedite the advancement of AD combinational therapy through providing an exceptionally convenient platform. This article is protected by copyright. All rights reserved.

Selective Formation of Homochiral Dimers by Intermolecular Charge Transfer on a hBN Nanomesh.

In this study, a comprehensive characterization was conducted on a chiral starburst molecule (C57H48N4, SBM) using scanning tunneling microscopy. When adsorbed onto the hBN/Rh(111) nanomesh, these molecules demonstrate homochiral recognition, leading to a selective formation of homochiral dimers. Further tip manipulation experiments reveal that the chiral dimers are stable and primarily controlled by strong intermolecular interactions. Density functional theory (DFT) calculations supported that the chiral recognition of SBM molecules is governed by the intermolecular charge transfer mechanism, different from the common steric hindrance effect. This study emphasizes the importance of intermolecular charge transfer interactions, offering valuable insights into the chiral recognition of a simple bimolecular system. These findings hold significance for the future advancement in chirality-based electronic sensors and pharmaceuticals, where the chirality of molecules can impact their properties.

Mitochondrial dysfunction, a weakest link of network of aging, relation to innate intramitochondrial immunity of DNA recognition receptors.

Aging probably is the most complexed process in biology. It is manifested by a variety of hallmarks. These hallmarks weave a network of aging; however, each hallmark is not uniformly strong for the network. It is the weakest link determining the strengthening of the network of aging, or the maximum lifespan of an organism. Therefore, only improvement of the weakest link has the chance to increase the maximum lifespan but not others. We hypothesize that mitochondrial dysfunction is the weakest link of the network of aging. It may origin from the innate intramitochondrial immunity related to the activity of pathogen DNA recognition receptors. These receptors recognize mtDNA as the PAMP or DAMP to initiate the immune or inflammatory reactions. Evidence has shown that several of these receptors including TLR9, cGAS and IFI16 can be translocated into mitochondria. The potentially intramitochondrial pathogen DNA recognition receptors have the capacity to attack the exposed second structures of the mtDNA during its transcriptional or especially the replication processed, leading to the mtDNA mutation, deletion, heteroplasmy colonization, mitochondrial dysfunction, and alterations of other hallmarks, as well as aging. Pre-consumption of the intramitochondrial pathogen DNA recognition receptors by medical interventions including development of mitochondrial targeted small molecule which can neutralization of these receptors may retard or even reverse the aging to significantly improve the maximum lifespan of the organisms.

Predictive performance of NEWS and qSOFA in immunocompromised sepsis patients at the emergency department.

PURPOSE: Sepsis has a high incidence and a poor prognosis. Early recognition is important to facilitate timely initiation of adequate care. Sepsis screening tools, such as the (quick) Sequential Organ Failure Assessment ((q)SOFA) and National Early Warning Score (NEWS), could help recognize sepsis. These tools have been validated in a general immunocompetent population, while their performance in immunocompromised patients, who are particularly at risk of sepsis development, remains unknown. METHODS: This study is a post hoc analysis of a prospective observational study performed at the emergency department. Inclusion criteria were age ≥ 18 years with a suspected infection, while ≥ two qSOFA and/or SOFA criteria were used to classify patients as having suspected sepsis. The primary outcome was in-hospital mortality. RESULTS: 1516 patients, of which 40.5% used one or more immunosuppressives, were included. NEWS had a higher prognostic accuracy as compared to qSOFA for predicting poor outcome among immunocompromised sepsis patients. Of all tested immunosuppressives, high-dose glucocorticoid therapy was associated with a threefold increased risk of both in-hospital and 28-day mortality. CONCLUSION: In contrast to NEWS, qSOFA underestimates the risk of adverse outcome in patients using high-dose glucocorticoids. As a clinical consequence, to adequately assess the severity of illness among immunocompromised patients, health care professionals should best use the NEWS.

Type 1 secretion necessitates a tight interplay between all domains of the ABC transporter.

Type I secretion systems (T1SS) facilitate the secretion of substrates in one step across both membranes of Gram-negative bacteria. A prime example is the hemolysin T1SS which secretes the toxin HlyA. Secretion is energized by the ABC transporter HlyB, which forms a complex together with the membrane fusion protein HlyD and the outer membrane protein TolC. HlyB features three domains: an N-terminal C39 peptidase-like domain (CLD), a transmembrane domain (TMD) and a C-terminal nucleotide binding domain (NBD). Here, we created chimeric transporters by swapping one or more domains of HlyB with the respective domain(s) of RtxB, a HlyB homolog from Kingella kingae. We tested all chimeric transporters for their ability to secrete pro-HlyA when co-expressed with HlyD. The CLD proved to be most critical, as a substitution abolished secretion. Swapping only the TMD or NBD reduced the secretion efficiency, while a simultaneous exchange abolished secretion. These results indicate that the CLD is the most critical secretion determinant, while TMD and NBD might possess additional recognition or interaction sites. This mode of recognition represents a hierarchical and extreme unusual case of substrate recognition for ABC transporters and optimal secretion requires a tight interplay between all domains.

Implications of depressive mood in OSAHS patients: insights from event-related potential.

BACKGROUND: Obstructive sleep apnea-hypopnea syndrome (OSAHS) is a chronic breathing disorder characterized by recurrent upper airway obstruction during sleep. Although previous studies have shown a link between OSAHS and depressive mood, the neurobiological mechanisms underlying mood disorders in OSAHS patients remain poorly understood. This study aims to investigate the emotion processing mechanism in OSAHS patients with depressive mood using event-related potentials (ERPs). METHODS: Seventy-four OSAHS patients were divided into the depressive mood and non-depressive mood groups according to their Self-rating Depression Scale (SDS) scores. Patients underwent overnight polysomnography and completed various cognitive and emotional questionnaires. The patients were shown facial images displaying positive, neutral, and negative emotions and tasked to identify the emotion category, while their visual evoked potential was simultaneously recorded. RESULTS: The two groups did not differ significantly in age, BMI, and years of education, but showed significant differences in their slow wave sleep ratio (P = 0.039), ESS (P = 0.006), MMSE (P < 0.001), and MOCA scores (P = 0.043). No significant difference was found in accuracy and response time on emotional face recognition between the two groups. N170 latency in the depressive group was significantly longer than the non-depressive group (P = 0.014 and 0.007) at the bilateral parieto-occipital lobe, while no significant difference in N170 amplitude was found. No significant difference in P300 amplitude or latency between the two groups. Furthermore, N170 amplitude at PO7 was positively correlated with the arousal index and negatively with MOCA scores (both P < 0.01). CONCLUSION: OSAHS patients with depressive mood exhibit increased N170 latency and impaired facial emotion recognition ability. Special attention towards the depressive mood among OSAHS patients is warranted for its implications for patient care.

Highly diverse cuticular hydrocarbon profiles but no evidence for aggression towards non-kin in the ambrosia beetle Xyleborinus saxesenii.

Animal societies use nestmate recognition to protect against social cheaters and parasites. In most social insect societies, individuals recognize and exclude any non-nestmates and the roles of cuticular hydrocarbons as recognition cues are well documented. Some ambrosia beetles live in cooperatively breeding societies with farmed fungus cultures that are challenging to establish, but of very high value once established. Hence, social cheaters that sneak into a nest without paying the costs of nest foundation may be selected. Therefore, nestmate recognition is also expected to exist in ambrosia beetles, but so far nobody has investigated this behavior and its underlying mechanisms. Here we studied the ability for nestmate recognition in the cooperatively breeding ambrosia beetle Xyleborinus saxesenii, combining behavioural observations and cuticular hydrocarbon analyses. Laboratory nests of X. saxesenii were exposed to foreign adult females from the same population, another population and another species. Survival as well as the behaviours of the foreign female were observed. The behaviours of the receiving individuals were also observed. We expected that increasing genetic distance would cause increasing distance in chemical profiles and increasing levels of behavioural exclusion and possibly mortality. Chemical profiles differed between populations and appeared as variable as in other highly social insects. However, we found only very little evidence for the behavioural exclusion of foreign individuals. Interpopulation donors left nests at a higher rate than control donors, but neither their behaviours nor the behaviours of receiver individuals within the nest showed any response to the foreign individual in either of the treatments. These results suggest that cuticular hydrocarbon profiles might be used for communication and nestmate recognition, but that behavioural exclusion of non-nestmates is either absent in X. saxesenii or that agonistic encounters are so rare or subtle that they could not be detected by our method. Additional studies are needed to investigate this further.

Understanding of facial features in face perception: insights from deep convolutional neural networks.

Introduction: Face recognition has been a longstanding subject of interest in the fields of cognitive neuroscience and computer vision research. One key focus has been to understand the relative importance of different facial features in identifying individuals. Previous studies in humans have demonstrated the crucial role of eyebrows in face recognition, potentially even surpassing the importance of the eyes. However, eyebrows are not only vital for face recognition but also play a significant role in recognizing facial expressions and intentions, which might occur simultaneously and influence the face recognition process. Methods: To address these challenges, our current study aimed to leverage the power of deep convolutional neural networks (DCNNs), an artificial face recognition system, which can be specifically tailored for face recognition tasks. In this study, we investigated the relative importance of various facial features in face recognition by selectively blocking feature information from the input to the DCNN. Additionally, we conducted experiments in which we systematically blurred the information related to eyebrows to varying degrees. Results: Our findings aligned with previous human research, revealing that eyebrows are the most critical feature for face recognition, followed by eyes, mouth, and nose, in that order. The results demonstrated that the presence of eyebrows was more crucial than their specific high-frequency details, such as edges and textures, compared to other facial features, where the details also played a significant role. Furthermore, our results revealed that, unlike other facial features, the activation map indicated that the significance of eyebrows areas could not be readily adjusted to compensate for the absence of eyebrow information. This finding explains why masking eyebrows led to more significant deficits in face recognition performance. Additionally, we observed a synergistic relationship among facial features, providing evidence for holistic processing of faces within the DCNN. Discussion: Overall, our study sheds light on the underlying mechanisms of face recognition and underscores the potential of using DCNNs as valuable tools for further exploration in this field.

An improved corner dealiasing and recognition algorithm for 2D Wadell roundness computation.

This paper optimizes the 2D Wadell roundness calculation of particles based on digital image processing methods. An algorithm for grouping corner key points is proposed to distinguish each independent corner. Additionally, the cyclic midpoint filtering method is introduced for corner dealiasing, aiming to mitigate aliasing issues effectively. The relationships between the number of corner pixels (m), the central angle of the corner (α) and the parameter of the dealiasing degree (n) are established. The Krumbein chart and a sandstone thin section image were used as examples to calculate the 2D Wadell roundness. A set of regular shapes is calculated, and the error of this method is discussed. When α ≥ 30°, the maximum error of Wadell roundness for regular shapes is 5.21%; when 12° ≤ α < 30°, the maximum error increases. By applying interpolation to increase the corner pixels to the minimum number (m0) within the allowable range of error, based on the α-m0 relational expression obtained in this study, the error of the corner circle can be minimized. The results indicate that as the value of m increases, the optimal range interval for n also widens. Additionally, a higher value of α leads to a lower dependence on m. The study's results can be applied to dealiasing and shape analysis of complex closed contours.

Using deep learning-based artificial intelligence electronic images in improving middle school teachers' literacy.

With the rapid development of societal information, electronic educational resources have become an indispensable component of modern education. In response to the increasingly formidable challenges faced by secondary school teachers, this study endeavors to analyze and explore the application of artificial intelligence (AI) methods to enhance their cognitive literacy. Initially, this discourse delves into the application of AI-generated electronic images in the training and instruction of middle school educators, subjecting it to thorough analysis. Emphasis is placed on elucidating the pivotal role played by AI electronic images in elevating the proficiency of middle school teachers. Subsequently, an integrated intelligent device serves as the foundation for establishing a model that applies intelligent classification and algorithms based on the Structure of the Observed Learning Outcome (SOLO). This model is designed to assess the cognitive literacy and teaching efficacy of middle school educators, and its performance is juxtaposed with classification algorithms such as support vector machine (SVM) and decision trees. The findings reveal that, following 600 iterations of the model, the SVM algorithm achieves a 77% accuracy rate in recognizing teacher literacy, whereas the SOLO algorithm attains 80%. Concurrently, the spatial complexities of the SVM-based and SOLO-based intelligent literacy improvement models are determined to be 45 and 22, respectively. Notably, it is discerned that, with escalating iterations, the SOLO algorithm exhibits higher accuracy and reduced spatial complexity in evaluating teachers' pedagogical literacy. Consequently, the utilization of AI methodologies proves highly efficacious in advancing electronic imaging technology and enhancing the efficacy of image recognition in educational instruction.

A novel approach to texture recognition combining deep learning orthogonal convolution with regional input features.

Textures provide a powerful segmentation and object detection cue. Recent research has shown that deep convolutional nets like Visual Geometry Group (VGG) and ResNet perform well in non-stationary texture datasets. Non-stationary textures have local structures that change from one region of the image to the other. This is consistent with the view that deep convolutional networks are good at detecting local microstructures disguised as textures. However, stationary textures are textures that have statistical properties that are constant or slow varying over the entire region are not well detected by deep convolutional networks. This research demonstrates that simple seven-layer convolutional networks can obtain better results than deep networks using a novel convolutional technique called orthogonal convolution with pre-calculated regional features using grey level co-occurrence matrix. We obtained an average of 8.5% improvement in accuracy in texture recognition on the Outex dataset over GoogleNet, ResNet, VGG and AlexNet.

E-MFNN: an emotion-multimodal fusion neural network framework for emotion recognition.

Emotional recognition is a pivotal research domain in computer and cognitive science. Recent advancements have led to various emotion recognition methods, leveraging data from diverse sources like speech, facial expressions, electroencephalogram (EEG), electrocardiogram, and eye tracking (ET). This article introduces a novel emotion recognition framework, primarily targeting the analysis of users' psychological reactions and stimuli. It is important to note that the stimuli eliciting emotional responses are as critical as the responses themselves. Hence, our approach synergizes stimulus data with physical and physiological signals, pioneering a multimodal method for emotional cognition. Our proposed framework unites stimulus source data with physiological signals, aiming to enhance the accuracy and robustness of emotion recognition through data integration. We initiated an emotional cognition experiment to gather EEG and ET data alongside recording emotional responses. Building on this, we developed the Emotion-Multimodal Fusion Neural Network (E-MFNN), optimized for multimodal data fusion to process both stimulus and physiological data. We conducted extensive comparisons between our framework's outcomes and those from existing models, also assessing various algorithmic approaches within our framework. This comparison underscores our framework's efficacy in multimodal emotion recognition. The source code is publicly available at https://figshare.com/s/8833d837871c78542b29.

Emotion detection from handwriting and drawing samples using an attention-based transformer model.

Emotion detection (ED) involves the identification and understanding of an individual's emotional state through various cues such as facial expressions, voice tones, physiological changes, and behavioral patterns. In this context, behavioral analysis is employed to observe actions and behaviors for emotional interpretation. This work specifically employs behavioral metrics like drawing and handwriting to determine a person's emotional state, recognizing these actions as physical functions integrating motor and cognitive processes. The study proposes an attention-based transformer model as an innovative approach to identify emotions from handwriting and drawing samples, thereby advancing the capabilities of ED into the domains of fine motor skills and artistic expression. The initial data obtained provides a set of points that correspond to the handwriting or drawing strokes. Each stroke point is subsequently delivered to the attention-based transformer model, which embeds it into a high-dimensional vector space. The model builds a prediction about the emotional state of the person who generated the sample by integrating the most important components and patterns in the input sequence using self-attentional processes. The proposed approach possesses a distinct advantage in its enhanced capacity to capture long-range correlations compared to conventional recurrent neural networks (RNN). This characteristic makes it particularly well-suited for the precise identification of emotions from samples of handwriting and drawings, signifying a notable advancement in the field of emotion detection. The proposed method produced cutting-edge outcomes of 92.64% on the benchmark dataset known as EMOTHAW (Emotion Recognition via Handwriting and Drawing).

Customized deep learning based Turkish automatic speech recognition system supported by language model.

Background: In today's world, numerous applications integral to various facets of daily life include automatic speech recognition methods. Thus, the development of a successful automatic speech recognition system can significantly augment the convenience of people's daily routines. While many automatic speech recognition systems have been established for widely spoken languages like English, there has been insufficient progress in developing such systems for less common languages such as Turkish. Moreover, due to its agglutinative structure, designing a speech recognition system for Turkish presents greater challenges compared to other language groups. Therefore, our study focused on proposing deep learning models for automatic speech recognition in Turkish, complemented by the integration of a language model. Methods: In our study, deep learning models were formulated by incorporating convolutional neural networks, gated recurrent units, long short-term memories, and transformer layers. The Zemberek library was employed to craft the language model to improve system performance. Furthermore, the Bayesian optimization method was applied to fine-tune the hyper-parameters of the deep learning models. To evaluate the model's performance, standard metrics widely used in automatic speech recognition systems, specifically word error rate and character error rate scores, were employed. Results: Upon reviewing the experimental results, it becomes evident that when optimal hyper-parameters are applied to models developed with various layers, the scores are as follows: Without the use of a language model, the Turkish Microphone Speech Corpus dataset yields scores of 22.2 -word error rate and 14.05-character error rate, while the Turkish Speech Corpus dataset results in scores of 11.5 -word error rate and 4.15 character error rate. Upon incorporating the language model, notable improvements were observed. Specifically, for the Turkish Microphone Speech Corpus dataset, the word error rate score decreased to 9.85, and the character error rate score lowered to 5.35. Similarly, the word error rate score improved to 8.4, and the character error rate score decreased to 2.7 for the Turkish Speech Corpus dataset. These results demonstrate that our model outperforms the studies found in the existing literature.

Deep learning-based recognition system for pashto handwritten text: benchmark on PHTI.

This article introduces a recognition system for handwritten text in the Pashto language, representing the first attempt to establish a baseline system using the Pashto Handwritten Text Imagebase (PHTI) dataset. Initially, the PHTI dataset underwent pre-processed to eliminate unwanted characters, subsequently, the dataset was divided into training 70%, validation 15%, and test sets 15%. The proposed recognition system is based on multi-dimensional long short-term memory (MD-LSTM) networks. A comprehensive empirical analysis was conducted to determine the optimal parameters for the proposed MD-LSTM architecture; Counter experiments were used to evaluate the performance of the proposed system comparing with the state-of-the-art models on the PHTI dataset. The novelty of our proposed model, compared to other state of the art models, lies in its hidden layer size (i.e., 10, 20, 80) and its Tanh layer size (i.e., 20, 40). The system achieves a Character Error Rate (CER) of 20.77% as a baseline on the test set. The top 20 confusions are reported to check the performance and limitations of the proposed model. The results highlight complications and future perspective of the Pashto language towards the digital transition.

Your blush gives you away: detecting hidden mental states with remote photoplethysmography and thermal imaging.

Multimodal emotion recognition techniques are increasingly essential for assessing mental states. Image-based methods, however, tend to focus predominantly on overt visual cues and often overlook subtler mental state changes. Psychophysiological research has demonstrated that heart rate (HR) and skin temperature are effective in detecting autonomic nervous system (ANS) activities, thereby revealing these subtle changes. However, traditional HR tools are generally more costly and less portable, while skin temperature analysis usually necessitates extensive manual processing. Advances in remote photoplethysmography (r-PPG) and automatic thermal region of interest (ROI) detection algorithms have been developed to address these issues, yet their accuracy in practical applications remains limited. This study aims to bridge this gap by integrating r-PPG with thermal imaging to enhance prediction performance. Ninety participants completed a 20-min questionnaire to induce cognitive stress, followed by watching a film aimed at eliciting moral elevation. The results demonstrate that the combination of r-PPG and thermal imaging effectively detects emotional shifts. Using r-PPG alone, the prediction accuracy was 77% for cognitive stress and 61% for moral elevation, as determined by a support vector machine (SVM). Thermal imaging alone achieved 79% accuracy for cognitive stress and 78% for moral elevation, utilizing a random forest (RF) algorithm. An early fusion strategy of these modalities significantly improved accuracies, achieving 87% for cognitive stress and 83% for moral elevation using RF. Further analysis, which utilized statistical metrics and explainable machine learning methods including SHapley Additive exPlanations (SHAP), highlighted key features and clarified the relationship between cardiac responses and facial temperature variations. Notably, it was observed that cardiovascular features derived from r-PPG models had a more pronounced influence in data fusion, despite thermal imaging's higher predictive accuracy in unimodal analysis.

YOLOv8s-CGF: a lightweight model for wheat ear Fusarium head blight detection.

Fusarium head blight (FHB) is a destructive disease that affects wheat production. Detecting FHB accurately and rapidly is crucial for improving wheat yield. Traditional models are difficult to apply to mobile devices due to large parameters, high computation, and resource requirements. Therefore, this article proposes a lightweight detection method based on an improved YOLOv8s to facilitate the rapid deployment of the model on mobile terminals and improve the detection efficiency of wheat FHB. The proposed method introduced a C-FasterNet module, which replaced the C2f module in the backbone network. It helps reduce the number of parameters and the computational volume of the model. Additionally, the Conv in the backbone network is replaced with GhostConv, further reducing parameters and computation without significantly affecting detection accuracy. Thirdly, the introduction of the Focal CIoU loss function reduces the impact of sample imbalance on the detection results and accelerates the model convergence. Lastly, the large target detection head was removed from the model for lightweight. The experimental results show that the size of the improved model (YOLOv8s-CGF) is only 11.7 M, which accounts for 52.0% of the original model (YOLOv8s). The number of parameters is only 5.7 × 106 M, equivalent to 51.4% of the original model. The computational volume is only 21.1 GFLOPs, representing 74.3% of the original model. Moreover, the mean average precision (mAP@0.5) of the model is 99.492%, which is 0.003% higher than the original model, and the mAP@0.5:0.95 is 0.269% higher than the original model. Compared to other YOLO models, the improved lightweight model not only achieved the highest detection precision but also significantly reduced the number of parameters and model size. This provides a valuable reference for FHB detection in wheat ears and deployment on mobile terminals in field environments.