Hao Wang

Improving uneven exposure using color characteristics as a priori information in endoscopic images

Yan Wang, Hao Wang, Xiaopan Xu, Kun Yang et autres

cn (code pays fourni par la source)

Integrated generation-detection approach for pedestrian recognition in dark-light conditions

Yuxiang Shen, Xin Guo, Hao Wang, Kai Hu et autres

目的暗光环境下的行人检测是计算机视觉领域的一项重大挑战，其难点在于光照不充分会导致行人特征模糊不可辨识。传统方法一般通过图像增强或生成红外图像提供补充信息解决这一问题。但由于其增强过程与下游检测任务存在分离，限制其应用性能。针对这一问题，提出一种面向暗光条件下行人识别的生成检测一体化方法，旨在解决传统范式中暗光图像数据增强和下游检测任务间的割裂问题。方法提出一种端到端生成检测一体化架构，利用条件扩散模型从暗光图像生成辅助红外模态图像，并通过生成的多尺度特征提升目标检测性能。此外，为了避免梯度无法经过VAE（variational autoencoder） 解码器传递的问题，进一步提出生成检测端到端联合优化策略。结果在LLVIP（low-light visible-infrared paired dataset）和VTMOT（visible-thermal multiple object tracking dataset）数据集上的实验结果表明，本文方法在总体精度指标上显著优于传统方法和其他先进方法。在LLVIP数据集上，本文方法的F1值为85.75%，平均精度均值（mAP）为75.35%，优于传统方法中结果最佳的SCI（self-calibrated illumination）方法（F1值82.05%，mAP值72.30%）和其他先进方法中结果最佳的Faster_RCNN_hrnet（F1值82.91%，mAP值70.02%）。在VTMOT数据集上，本文方法同样表现优异，F1值为90.01%，mAP为73.44%，优于SCI方法（F1值85.91%，mAP值72.19%）和Faster_RCNN_hrnet（F1值84.48%，mAP值71.16%）。此外，消融实验验证了生成模块和联合优化策略在整体框架中的有效性。结论本文证明了生成检测一体化框架在复杂低光环境下的优越性，有效解决了生成过程与检测任务割裂的问题。未来研究将进一步优化生成效率，并扩展该方法至更多多模态应用领域。

cn (code pays fourni par la source)

Generative Model-Based Edge-Assisted Object Detection in Bandwidth-Constrained Network

Jiayuan Liu, Ke Luo, Hao Wang, Xu Chen

cn (code pays fourni par la source)

Detecting Artificial Intelligence-Generated images via deep trace representations and interactive feature fusion

Qiang Xu, Xinghao Jiang, Tanfeng Sun, Hao Wang et autres

cn, hk (code pays fourni par la source)

Multi-Scale Feature Extraction and Monocular Depth Estimation Algorithm for Traffic Images

Hao Wang, Pengyu Liu

Unsupervised monocular depth estimation, also known as self-supervised monocular depth estimation, predicts the depth information of each pixel in a scene from unlabelled images or videos captured by a single camera, without requiring any manually annotated depth data. This avoids the complexity …

cn (code pays fourni par la source)

Exposing fake images generated by text-to-image diffusion models

Qiang Xu, Hao Wang, Laijin Meng, Zhongjie Mi et autres

hk, cn (code pays fourni par la source)

Semi-Supervised Wide-Angle Portraits Correction by Multi-Scale Transformer

Fushun Zhu, Shan Zhao, Peng Wang, Hao Wang et autres

We propose a semi-supervised network for wide-angle portraits correction. Wide-angle images often suffer from skew and distortion affected by perspective distortion, especially noticeable at the face regions. Previous deep learning based approaches need the ground-truth correction flow maps for training guidance. However, …

Research on image classification based on improved DenseNet

Hao Wang, Song Zhili

In order to improve the accuracy of image classification, many researchers will improve the network structure, enhance the way of data processing or design a new activation function. This paper improves the initial convolution block and activation function of DenseNet's network structure, …

cn (code pays fourni par la source)

Deep High-Resolution Representation Learning for Cross-Resolution Person Re-identification

Guoqing Zhang, Yu Ge, Zhicheng Dong, Hao Wang et autres

Person re-identification (re-ID) tackles the problem of matching person images with the same identity from different cameras. In practical applications, due to the differences in camera performance and distance between cameras and persons of interest, captured person images usually have various resolutions. …

cn (code pays fourni par la source)

Medical Image Registration Based on Uncoupled Learning and Accumulative Enhancement

Yucheng Shu, Hao Wang, Bin Xiao, Xiuli Bi et autres

cn (code pays fourni par la source)

Learning Guided Electron Microscopy with Active Acquisition

Mi Lu, Hao Wang, Yaron Meirovitch, Richard Schalek et autres

us (code pays fourni par la source)

Task-agnostic Temporally Consistent Facial Video Editing

Meng Cao, Haozhi Huang, Hao Wang, Xuan Wang et autres

Recent research has witnessed the advances in facial image editing tasks. For video editing, however, previous methods either simply apply transformations frame by frame or utilize multiple frames in a concatenated or iterative fashion, which leads to noticeable visual flickers. In addition, …