With the rapid advancements of the text-to-image generative model, AI-generated images (AGIs) have been widely applied to entertainment, education, social media, etc. However, considering the large quality variance among different AGIs, there is an urgent need for quality models that are consistent with human subjective ratings. To address this issue, we extensively consider various popular AGI models, generated AGI through different prompts and model parameters, and collected subjective scores at the perceptual quality and text-to-image alignment, thus building the most comprehensive AGI subjective quality database AGIQA-3K so far. Furthermore, we conduct a benchmark experiment on this database to evaluate the consistency between the current Image Quality Assessment (IQA) model and human perception, while proposing StairReward that significantly improves the assessment performance of subjective text-to-image alignment. We believe that the fine-grained subjective scores in AGIQA-3K will inspire subsequent AGI quality models to fit human subjective perception mechanisms at both perception and alignment levels and to optimize the generation result of future AGI models. The database is released on https://github.com/lcysyzxdxc/AGIQA-3k-Database.

AGIQA-3K: An Open Database for AI-Generated Image Quality Assessment

Digital humans have witnessed extensive applications in various domains, necessitating related quality assessment studies. However, there is a lack of comprehensive digital human quality assessment (DHQA) databases. To address this gap, we propose SJTU-H3D, a subjective quality assessment database specifically designed for full-body digital humans. It comprises 40 high-quality reference digital humans and 1,120 labeled distorted counterparts generated with seven types of distortions. The SJTU-H3D database can serve as a benchmark for DHQA research, allowing evaluation and refinement of processing algorithms. Further, we propose a zero-shot DHQA approach that focuses on no-reference (NR) scenarios to ensure generalization capabilities while mitigating database bias. Our method leverages semantic and distortion features extracted from projections, as well as geometry features derived from the mesh structure of digital humans. Specifically, we employ the Contrastive Language-Image Pre-training (CLIP) model to measure semantic affinity and incorporate the Naturalness Image Quality Evaluator (NIQE) model to capture low-level distortion information. Additionally, we utilize dihedral angles as geometry descriptors to extract mesh features. By aggregating these measures, we introduce the Digital Human Quality Index (DHQI), which demonstrates significant improvements in zero-shot performance. The DHQI can also serve as a robust baseline for DHQA tasks, facilitating advancements in the field. The database and the code are available at https://github.com/zzc-1998/SJTU-H3D.

Advancing Zero-Shot Digital Human Quality Assessment through Text-Prompted Evaluation

Alzheimer's disease (AD) has become a major public health concern, and reliable screening and diagnosis remain difficult. Magnetic resonance image (MRI) can be helpful in distinguishing AD patients from individuals with normal cognition (NC). Deep neural networks have demonstrated strong capacities for extracting intricate nonlinear correlations from brain imaging data recently. However, this requires large amounts of data for training to avoid overfitting problems, but data is scarce and precious in medical field. In this work, we propose a Vision Transformer network architecture called Aux-ViT, which solves the problem of losing shallow features by adding a class auxiliary branch. Specifically, we choose ViT as the backbone network and add an Auxiliary Multi-layer Perceptron Head to output auxiliary prediction results for calculating prediction errors. Based on the characteristics of MRI, we also developed a brain MRI data preprocessing method called Multi-Information Fusion Improvement, while further achieving data enhancement using a random synthetic mask based on pixel weighting fusion. We conducted extensive experiments using the ADNI-3 dataset to validate our algorithm. When compared to the baseline ViT model, the Aux-ViT model obtains an accuracy of 89.58%, which is an increase in accuracy of 3.93% and a decrease in training time of 47.7%. Our study provides a practical approach for early Alzheimer's diagnostics utilizing MRI data.

Aux-ViT : Classification of Alzheimer's Disease from MRI based on Vision Transformer with Auxiliary Branch

Abstract Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is primarily transmitted from person to person through respiratory droplets and aerosols. It is also possible for the virus to be transmitted indirectly through environmental contamination. The likelihood of environmental transmission depends on several factors, including the survival time of the virus in respiratory secretions. However, the stability of SARS-CoV-2 in respiratory secretions has not been investigated. In this study, we compared the half-life of the SARS-CoV-2 antigen in respiratory secretion under different conditions. We applied respiratory secretion (5 µL) to glass slides, air-dried the slides for 1 h, and kept them at 24 °C or 4 °C for 10 days. Respiratory secretions were also placed in test tubes (sealed to preserve moisture) and in normal saline for 10 days. The concentration of SARS-CoV-2 antigen in all samples was simultaneously measured using colloidal gold immunochromatography, and the half-life of the antigen was calculated. The half-life of the antigen in the wet (sealed tube) and saline samples at room temperature was 5.0 and 2.92 days, respectively. The half-life of the antigen in the air-dried sample at room temperature and at 4 °C was 2.93 and 11.4 days, respectively. The half-life was longer in respiratory secretions than that in normal saline. The half-life was also longer in respiratory secretions, at a lower temperature, and under wet conditions. Therefore, environmental transmission can also play a significant role in the spread of the virus. Robust prevention and control strategies could be developed based on the half-life of the antigen in respiratory secretions. 

/pdf/determining-half-life-of-sars-cov-2-antigen-in-respiratory-j7gfyu0o.pdf

Determining half-life of SARS-CoV-2 antigen in respiratory secretion

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10011043

LightR-YOLOv5: A compact rotating detector for SARS-CoV-2 antigen-detection rapid diagnostic test results

Recently, Users Generated Content (UGC) videos becomes ubiquitous in our daily lives. However, due to the limitations of photographic equipments and techniques, UGC videos often contain various degradations, in which one of the most visually unfavorable effects is the underexposure. Therefore, corresponding video enhancement algorithms such as Low-Light Video Enhancement (LLVE) have been proposed to deal with the specific degradation. However, different from video enhancement algorithms, almost all existing Video Quality Assessment (VQA) models are built generally rather than specifically, which measure the quality of a video from a comprehensive perspective. To the best of our knowledge, there is no VQA model specially designed for videos enhanced by LLVE algorithms. To this end, we first construct a Low-Light Video Enhancement Quality Assessment (LLVE-QA) dataset in which 254 original low-light videos are collected and then enhanced by leveraging 8 LLVE algorithms to obtain 2,060 videos in total. Moreover, we propose a quality assessment model specialized in LLVE, named Light-VQA. More concretely, since the brightness and noise have the most impact on low-light enhanced VQA, we handcraft corresponding features and integrate them with deep-learning-based semantic features as the overall spatial information. As for temporal information, in addition to deep-learning-based motion features, we also investigate the handcrafted brightness consistency among video frames, and the overall temporal information is their concatenation. Subsequently, spatial and temporal information is fused to obtain the quality-aware representation of a video. Extensive experimental results show that our Light-VQA achieves the best performance against the current State-Of-The-Art (SOTA) on LLVE-QA and public dataset. Dataset and Codes can be found at https://github.com/wenzhouyidu/Light-VQA.

Tao Tan

Papers

LightR-YOLOv5: A compact rotating detector for SARS-CoV-2 antigen-detection rapid diagnostic test results

Light-VQA: A Multi-Dimensional Quality Assessment Model for Low-Light Video Enhancement

Q-Learning with Heterogeneous Update Strategy

Ultrasonic evaluation of fetal lung development using deep learning with graph

PCTMF-Net: heart sound classification with parallel CNNs-transformer and second-order spectral analysis