YOLOv3-tiny is widely used in edge-oriented object detection, but its deployment on resource-constrained platforms is limited by high computational cost and the limited flexibility of conventional processors. This paper presents a RISC-V-based acceleration framework for YOLOv3-tiny inference that combines a tightly coupled CPU-accelerator architecture with runtime-reconfigurable hardware support. A Hummingbird E203 core is integrated with a dedicated accelerator through the NICE interface, and 11 custom instructions are introduced for data movement, convolution control, and post-processing. The hardware adopts a runtime-reconfigurable systolic array supporting multiple convolution kernel sizes, together with activation, pooling, fully connected, and detection-oriented post-processing modules. The design is implemented on an Artix-7 FPGA and evaluated using a hardware-oriented YOLOv3-tiny workload, supplemented by module-level analysis and same-platform baseline comparisons. Experimental results show a 79.5% reduction in convolution execution time and a 4.89 × speed-up over the baseline RISC-V processor. Hardware-supported post-processing further reduces the cycle cost of sorting and IoU-related computation by 60.55% and 45.44%, respectively. These results demonstrate the effectiveness of the proposed processor-coupled acceleration architecture for YOLOv3-tiny-based detection inference on edge platforms.
Diabetic retinopathy (DR) is a leading cause of preventable blindness, which has motivated the development of reliable automated grading systems on retinal fundus images. In this study, we perform a controlled comparative evaluation of ConvNeXt-Tiny, Swin-Tiny and their feature fusion for DR classification using the Asia Pacific Tele-Ophthalmology Society (APTOS) 2019 dataset. All models were initialized with weights pre-trained on ImageNet-1K and evaluated with two transfer learning strategies: direct fine-tuning on APTOS 2019, and EyePACS-based domain adaptation with task-specific fine-tuning. Systematic ablation experiments were carried out to evaluate the contribution of Contrast Limited Adaptive Histogram Equalization (CLAHE) preprocessing and channel-spatial attention modules (CSAM). We carried out experiments on the APTOS 2019 dataset with fixed train, validation and test splits and evaluated model stability across three runs with different random seeds by reporting mean ± standard deviation of performance metrics, while performance varied widely across architectures and training settings. After domain adaptation, the fusion-based models achieved more balanced results, while the standalone Swin-Tiny showed weaker adaptation to the retinal imaging domain, and was less sensitive to subtle lesion patterns under the EyePACS-based transfer learning. Adding CLAHE preprocessing and CSAM integration did not consistently improve class-balanced metrics. The best fusion configuration achieved a mean test accuracy of 88.34% ± 1.09 and a macro F1-score of 0.7376 ± 0.0183 on the APTOS 2019 dataset across repeated runs. These results suggest that domain-specific adaptation and architectural complementarity are more beneficial in boosting DR classification performance than auxiliary preprocessing or attention enhancement. The study also emphasizes the importance of controlled comparative evaluation, stability analysis, and configuration-specific evaluation in the research of medical image classification.
To address the challenges of low detection precision, severe background interference, and high model complexity in tiny crop disease lesion detection (defined as lesions occupying 8×8 to 32×32 pixels at 640×640 input resolution) under complex field environments, this study proposes a lightweight detection model named FDA-YOLO by integrating frequency-domain attention and improved YOLOv11. The model employs EfficientFormerV2 as the backbone to extract multi-scale features with low computational cost, and introduces a frequency domain attention module to enhance high-frequency tiny disease lesion details and suppress background noise. Comprehensive experiments on the PlantDoc dataset demonstrate that the proposed model achieves 96.3% mAP@0.5, 96.8% precision, and 36.4 FPS with only 28.5M parameters, outperforming the selected baseline detectors under the adopted experimental setting. The model realizes an optimal balance between accuracy, efficiency, and lightweight performance, providing a reliable and practical solution for real-time tiny lesion detection inprecision agriculture and edge device deployment.
Edge-side acoustic monitoring enables animal sound recognition in remote environments, but microcontroller deployment remains constrained by feature extraction, numerical consistency, memory, latency, and energy consumption. This study presents a sensor-based tiny machine learning (TinyML) acoustic monitoring system on an Arduino Nano 33 BLE Sense Rev2 platform, integrating onboard pulse-density modulation (PDM) microphone acquisition, Mel-frequency cepstral coefficient (MFCC) feature extraction, deployment-side standardization, 8-bit integer (INT8) neural-network inference, and edge-side decision output. To reduce training-to-deployment feature drift, consistent frame parameters, mirrored C++ feature operators, and exported standardization parameters are used to align personal-computer-side and microcontroller-side feature representations. A source-isolated seven-class protocol was constructed for six target animal classes and one compound background-noise class. In the single-run baseline comparison, the proposed multilayer perceptron achieved 98.28% test accuracy and 97.21% test macro-F1, while the ten-seed stability analysis yielded 98.64% ± 0.26% test accuracy and 97.87% ± 0.38% test macro-F1. The deployed INT8 model occupied approximately 26.9 KB, with a post-window latency of about 303 ms. System-level input power was 0.783-0.825 W, corresponding to an estimated autonomy of 7.63-8.03 h under the reference battery setting.
Autism Spectrum Disorder (ASD) in toddlers is characterized by neurodevelopmental deficits while its early detection remains challenging due to the lack of specific biomarkers and developmental variability. Therefore, a TinyDINO VHaar based Bi-directional Factorization with Chameleon optimized LightASDNet (TiDI-ASDNet) is proposed. In this framework, input images are initially pre-processed using Optimized Hierarchical Guided Image Filter (OHGF) to denoise and outlier removal, while Self-Distillation with No Labels version 2-Network Vector of Locally Aggregated Descriptors (DINOv2-NetVLAD) extracts the global visual patterns. Simultaneously, the input texts are pre-processed using One-Hot SMOTE (OHS) for easier interpretation. Besides, the standard structures interrupt syntactic/structural parsing due to autistic language corpora which clusters n-gram pattern, weakening the pragmatic deficits. To address these issues a Hybrid Tiny Encoder based HaarNet (TE-HNet) that extracts discriminative questionnaire features and maintains the lexical stringency. Since, the manifold torsion occurs due to discrepancy in encoding levels from non-isomorphic latent structures with non-diffeomorphic mapping, Bi-directional Encoder based Cross Factorization (BiE-xF) is employed and it learns about the shared semantics of visual features and behavioral linguistic features, which reduces Heteroscedastic Ambiguity. Besides, the motor stimming behaviour produces recurring self-stimulatory motor patterns that confound temporal alignment in vision-based ASD models. Thus, Modified Chameleon optimized LightASDNet (MC-LAN) is presented for classifying ASD and non-ASD thereby mitigates Dyspraxic Gait Aberrations. Simulations revealed the robustness of the framework with 99.2% AUC and 98.2% accuracy.
The outbreak of pests seriously affects the yield and quality of wheat crops. The accurate recognition and detection play an essential role in the early warning of crop pests. While some limitations, like insufficient dataset, imbalanced samples of pests with dense distribution, and dense distribution and tinyof crop pests, pose significant challenges to the precise detection. Thus, in this work, we first spent two years collecting real-world wheat pest images with four types of pests, including three grain aphids, and one mite species, to obtain a high-quality crop pest dataset for network optimization. Secondly, to alleviate insufficiency of samples of pests with dense distribution, we have developed a cut-up data augmentation strategy that separates dense pest targets from complex backgrounds. Furthermore, to address the challenge of pest detection with tiny body size and dense distribution, we introduce the Multiple Kernel Attention Network (MKA-Net), which further integrates the multi-scale features of pests to improve detection accuracy. Our method achieved the best detection precision, with AP50 reaching its peak at 67.1%, which is a significant improvement of nearly 6.9 points in wheat pest detection compared with the baseline. In summary, our proposed method can assist in the prevention and control of wheat pests and promote the progress of intelligent agriculture.
The presence of anthropogenic microparticles (particles <5 mm in any dimension) is being increasingly reported, yet little is known about their abundance, distribution, and bioavailability in marine environments. Microplastics (plastic polymers <5 mm) are persistent anthropogenic microparticles that contaminate marine environments and often enter food webs through ingestion. Sharks are sentinel species in the ocean, yet studies characterizing microparticle, or more specifically microplastic, ingestion in sharks are scarce. The objectives of this study were to describe and quantify microparticles in the feces of white sharks (Carcharodon carcharias). Samples were obtained from a total of 35 white sharks, including seven young-of-the-year, 17 juveniles, six subadults, and five adults inhabiting the Western North Atlantic Ocean. Microparticles were isolated, analyzed with a stereomicroscope to determine size, shape, and color, and characterized by polymer type using Fourier transform infrared (FTIR) spectroscopy. Anthropogenic microparticles were detected in all fecal samples. The mean ± standard deviation of microparticles per sample was 7.5 ± 4.2, and of those, 1.4 ± 1.3 were microplastics. No significant differences in the abundance of microplastics were detected by age class, sex, or collection location, year, or seasonal residency; however, differences in microparticle composition were observed across collection locations. Rayon and cellulose were the most abundant microparticle polymers, and polyester, polypropylene, and nylon were the most abundant microplastic polymers. White, black, and blue were the most common colors of microparticles and microplastics. This research contributes much-needed data from a top-level predator in marine ecosystems. Assessing microplastic content in white shark feces can serve as a non-lethal indicator of environmental contamination.
Movement is critical for animals' dynamics and survival. However, the movement patterns of small-bodied animals (body mass <500 g) are rarely documented, given the difficulties in tracking them in their natural habitats. Here, we report the distance travelled by a small-bodied and nocturnal lemur (brown mouse lemur, Microcebus rufus) in Ranomafana National Park, southeastern Madagascar, using telemetry and tracking data ( N = 7 night-tracking from 5 individuals). We found that M. rufus travelled up to 840.41 m in one night with a maximum 1-hour step length of 502.90 m. Our data also showed that, on average, females (755.50 m) travelled further than males (381.90 m). We suggest that such long-distance movements could be a strategy to maintain their metabolic rate and budget activities, avoid predators, and/or respond to changes in food source distribution. Furthermore, such movements can have critical consequences on their ecological roles in the rainforest, contributing to the self-maintenance of diversity, health, and regeneration of the ecosystem.
Telesurgery, the use of robotic systems to perform operations remotely, represents a growing frontier in minimally invasive surgery. We report a rare and pioneering case of a 16-month-old male with left pelviureteric junction obstruction. The child underwent robotic-assisted pyeloplasty performed via a telesurgical approach. The procedure was completed successfully without intraoperative or postoperative complications. The child recovered uneventfully. To our knowledge, this is among the youngest reported cases of robotic-assisted pyeloplasty performed via SSI Mantra Robotic system (Sudhir Srivastava Innovations Pvt. Ltd., India. It highlights the feasibility of telerobotic interventions in urology.
Precise identification of chili flowers and buds is essential for enabling intelligent robotic pollination, continuous crop monitoring, and early yield prediction in protected horticulture. Reliable visual sensing remains challenging because chili targets are extremely small, densely distributed, and frequently occluded by foliage, while greenhouse environments introduce strong illumination variations and background reflections. These factors often lead to insufficient feature extraction and unstable detection accuracy in existing models, limiting their practical deployment in automated monitoring systems. To address these challenges, this study proposes GCP-YOLO, a lightweight yet high-performance detection framework built upon the YOLOv11n architecture. The model enhances small-target perception through three key improvements. First, a redesigned Generalized Feature Pyramid Network (GFPN) strengthens cross-scale feature interaction, improving the fusion of fine-grained texture cues and deep semantic information. Second, a C2CGA context-guided attention module is introduced to emphasize floral structural features while suppressing background noise caused by reflections and canopy clutter. Third, extended multi-scale detection heads (P2-P6) incorporate broader contextual information to reduce missed detections and false positives in dense planting scenarios. Experimental results on a custom chili flower dataset show that the proposed method achieves 92.8% precision, 83.7% recall, 90.8% mAP50, and 72.7% mAP50-95, improving upon the YOLOv11n baseline by 2.1, 1.3, 3.9, and 6.6 percentage points, respectively. Deployment on an NVIDIA Jetson AGX Orin edge platform demonstrates real-time inference at 97.9 FPS, confirming its suitability for on-device phenotyping. Overall, the proposed approach significantly improves detection robustness under complex greenhouse conditions, providing an effective visual sensing methodology for automated crop monitoring and data-driven yield estimation.
Colliding molecules react only when their orientations and point of impact fall within a narrow range.
Accessory proteins of SARS-CoV-2 play crucial roles in viral pathogenesis, yet their structural properties remain elusive. ORF7b, a small accessory protein comprising only 43 amino acids, is widely assumed to parallel the structure-function relationships of its SARS-CoV ortholog based solely on sequence homology. In this study, we challenge this paradigm through direct physicochemical and structural characterization. Sequence analysis and electrostatic profiling reveal that the SARS-CoV-2 protein is a macromolecular polyanion with a net charge of -4 at neutral pH, featuring a diffuse negative surface that is highly responsive to pH changes. Complete 3D structures generated via ab initio modeling display a helical core flanked by two highly fluctuating, disordered termini. Residue Interaction Network (RIN) topology and Normal Mode Analysis (NMA) identified specific hinges governing these flexible extremities. Furthermore, the calculated dipole moment vector is tilted outward by 24°, misaligning with the central axis. Molecular dynamics simulations suggest that while the soluble structure is highly stable in water, it undergoes severe distortions and insufficient solvation within a membrane-mimetic environment. Thermodynamic association profiles and verified interactomic data from BioGRID reveal a strong propensity for ORF7b to participate in liquid-liquid phase transitions alongside human and viral partners. Taken together, these unique properties suggest that ORF7b operates as a dynamic peripheral membrane protein rather than a sedentary transmembrane component, providing a fresh framework for future therapeutic targeting. Overall, these in silico findings shift the current paradigm on ORF7b2 topology and provide a robust, physically grounded framework that identifies specific molecular priorities for future in vitro and in vivo validation.
Although magnetic compression anastomosis procedures are increasing in number, the procedure is still under-recognized in the pediatric patient population. Post-surgical obstructions encountered in biliary, gastrointestinal, and urinary tracts can be managed with magnetic compression anastomosis even in the pediatric population. Considering the challenges and comorbidities of revision surgery in this particular patient population, magnets can provide a less invasive and straightforward solution.
Gastric metastasis from clear cell renal cell carcinoma is uncommon and may mimic a primary gastric neoplasm. We report a man in his late sixties with a 0.6-cm reddish protruding gastric lesion initially interpreted as fundic gland-type adenocarcinoma, mainly on the basis of morphology and an initial immunohistochemical panel. Targeted forceps biopsy showed small nests of tumor cells in the lamina propria. Endoscopic submucosal dissection, performed under the working diagnosis of early primary gastric neoplasia, showed no residual malignancy, probably because the minute focus had been largely removed by biopsy. Subsequent nephrectomy confirmed clear cell renal cell carcinoma. Additional immunohistochemistry of the gastric biopsy showed PAX8, CAIX, CK8/18, partial CD10 and focal RCC marker positivity, supporting gastric metastasis. This case highlights the importance of renal-lineage immunohistochemistry in small gastric lesions with unusual morphology.
Drosophila melanogaster, widely known as the fruit fly, has emerged as a pivotal model organism for studying development and signaling transduction. Its fully sequenced genome, short generation time, and powerful genetic toolkit-including the Gal4/UAS system, RNA interference, and CRISPR-Cas9-enable precise, tissue-specific manipulation and high-throughput functional analyses. Despite differences in anatomy, the internal organ systems of Drosophila melanogaster, including the nervous system, heart, fat body, oenocytes, and nephrocytes, exhibit conserved molecular pathways and physiological functions comparable to those of humans. The morphological differences between invertebrates and vertebrates have long led researchers to undervalue the studies of insects in underlying the pathogenesis of human diseases. Over the past decades, the fruit fly has been widely validated for modeling the pathogenesis of neurodegenerative, cardiovascular, metabolic, renal, and muscular disorders. In this review, we systematically summarize the conserved molecular pathways and organ functions between the fruit fly and human, and provide examples of recent studies that use the fruit fly as a model system to answer questions associated with human diseases. We also discuss how Drosophila help researchers to fulfill the gap from mechanistic study toward translational research, and provide methodological considerations regarding the utility of Drosophila models in drug screening.
Traditional genetic transformation approaches relying on Agrobacterium tumefaciens for the delivery of CRISPR/Cas9 reagents usually provide plants that stably integrate the gene construct in their genome. To meet the EU commission's proposal for a new legislation on plants obtained by new genomic techniques (NGTs), it is important to develop new protocols that produce transgene-free genome edited plants (NGT category 1). Protoplasts are a promising platform, since delivery of CRISPR/Cas9 reagents as ribonucleoproteins (RNPs) is effective in cells lacking their wall. This allows genetic modifications from a transient application, leaving no traces in the recipient genome apart from the desired targeted mutations. With the aim of implementing transgene-free editing of eggplant (Solanum melongena L.), we adapted and improved a protocol previously established in potato and tomato for the isolation of protoplasts from cotyledonary leaves and subsequent CRISPR/Cas9 reagents delivery. Isolated protoplasts were subjected to in vitro culture and regeneration, and the first shoots were regenerated from calli approximately 4-5 months after isolation. Alongside, two transfection protocols were tested for the delivery of RNPs into eggplant protoplasts, one using polyethylene glycol (PEG) in two concentrations (25% and 40%) and one exploiting two formulations of lipofectamines (Lipofectamine CRISPRMAX™ and Lipofectamine™ 3000), all targeting SmChl_H gene, whose inactivation can cause a chlorotic phenotype. Efficient callus regeneration from transfected protoplasts was obtained and the editing efficiency (calculated as the percentage of edited calli on the total of calli that underwent sequencing) was evaluated. 25% PEG treatment provided the highest editing efficiency, and fully edited biallelic calli were retrieved, showing the expected chlorotic phenotype. Even if the efficiency of in vitro regeneration of plants from calli still needs improvement, edited plants were regenerated from protoplasts, representing the first report of RNP mediated genome editing in eggplant protoplasts.
Instruction-based image editing is a challenging task since it requires to manipulate the visual content of images according to complex human language instructions. When editing an image with tiny objects and complex positional relationships, existing image editing methods cannot locate the accurate region to execute the editing. To address this issue, we introduce Proximal Policy Optimization Guided Image Editing(PiGIE), a diffusion model that can accurately edit tiny objects in images with complex scenes. The PiGIE can incorporate proper noise masks to edit images based on the guidance of the target object's attention maps. Different from the traditional image editing approaches based on supervised learning, PiGIE leverages reinforcement learning with Proximal Policy Optimization (PPO) to fine-tune the diffusion model by using the cosine similarity between UNet attention maps and human feedback as the reward signal. On multiple image editing benchmarks, PiGIE exhibits remarkable improvements in both image quality and generalization capability. In particular, PiGIE sets a new baseline for editing fine-grained images with multiple tiny objects, shedding light on future studies on text-guided image editing for tiny objects.
To evaluate cross-population performance of deep learning models for referable diabetic retinopathy (DR) detection and assess whether inclusion of local data improves robustness in a Turkish external validation setting. Global datasets (DDR, IDRiD, Messidor) and a Turkish real-world clinical dataset were used. Images were re-graded by three ophthalmologists and binarized as referable/non-referable DR. Five architectures were evaluated: EfficientNet-B0, EfficientNet-V2, ResNet50, ConvNeXt-V2-Tiny, and Swin V2 CR Small. Models were tested under three settings: Global Train-Global Test, Global Train-Local Test, and Mixed Train-Mixed Test. Performance metrics included accuracy, sensitivity, specificity, NPV, ROC-AUC, and PR-AUC; Grad-CAM was used for explainability. The global training set included 9,881 images; two external test sets (2,042 images each) were used. Inter-grader agreement was high (ICC[3,1]=0.84; ICC[3,k]=0.96). Swin V2 CR Small achieved the highest performance in Global Train-Global Test (accuracy 0.938; ROC-AUC 0.985). Performance declined in Global Train-Local Test, where ConvNeXt-V2-Tiny performed best (accuracy 0.823; sensitivity 0.940; ROC-AUC 0.913). In Mixed Train-Mixed Test, ConvNeXt-V2-Tiny remained the most balanced model (accuracy 0.823; sensitivity 0.939; NPV 0.948; ROC-AUC 0.875). Models performed strongly under matched conditions but degraded across populations. ConvNeXt-V2-Tiny showed the most stable screening performance, supporting the need for local validation before deployment.
Students at two-year community and technical colleges need access to research opportunities that build scientific and interpersonal skills, and foster a scientific identity to successfully transfer to four-year institutions or the science, technology, engineering, and mathematics (STEM) workforce. Tiny Earth, a course-based undergraduate research experience (CURE) that advances antimicrobial discovery from soil bacteria, serves as the foundation for the Tiny Earth Summer Research Course (TESRC), a cohort-based, full-time summer research experience for undergraduates from historically underrepresented groups at two-year colleges to conduct research at a university after completing a Tiny Earth course at their enrolled institution. After completing TESRC, students reported gains in research and workforce skills, including understanding antimicrobial discovery, practicing the scientific method, communicating research findings, working collaboratively with peers and mentors, and solving problems independently. They also reported that participation in TESRC fostered a greater sense of belonging in a scientific community. After completing TESRC, students continued educational and career pursuits in scientific fields.