- 博客(26)
- 收藏
- 关注
原创 人体行为识别:Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset
参考文献:https://arxiv.org/abs/1705.07750pytorch代码实现:https://github.com/MRzzm/action-recognition-models-pytorchQuo Vadis, Action Recognition? A New Model and the Kinetics Dataset摘要由于目前动作分类数据集(UCF-101和HMDB-51)中视频的缺乏,大多数方法在小规模数据集基础上的性能相似,很难得到识别效果好的网络结构,本文根据最
2020-06-23 14:30:37
1629
原创 人体行为识别:Asynchronous Interaction Aggregation for Action Detection
参考文献:https://arxiv.org/abs/2004.07485v1代码实现:https://github.com/MVIG-SJTU/AlphActionAsynchronous Interaction Aggregation for Action Detection摘要理解交互是视频动作检测的重要组成部分。我们提出了异步交互聚合网络(AIA),它利用不同的交互促进动作检测。其中有两个关键设计:一是交互聚合结构(IA),采用统一的范式对多种交互类型进行建模和集成;另一种是异步内存更新算法
2020-06-20 17:45:41
4239
1
原创 人体行为识别:SlowFast Networks for Video Recognition
参考文献:https://arxiv.org/abs/1812.03982代码实现:https://github.com/facebookresearch/SlowFast包括理解!SlowFast Networks for Video Recognition摘要我们提出了用于视频识别的SlowFast网络,模型包括:(i)以低帧速率的慢速路径来捕获空间语义;(ii)以高帧速率的快速路径来捕获精细时间分辨率的运动。快速路径可以通过减少通道容量而变得非常轻量级,并且可以学习有用的时间信息用于视频识别
2020-06-18 17:14:39
6618
原创 EfficientDet: Scalable and Efficient Object Detection
参考文献:https://arxiv.org/pdf/1911.09070.pdfpytorch代码实现:https://github.com/zylo117/Yet-Another-EfficientDet-PytorchEfficientDet: Scalable and Efficient Object Detection摘要
2020-05-08 18:06:09
918
原创 可变形卷积网络:Deformable Convolutional Networks
参考文献:https://arxiv.org/abs/1703.06211代码实现:https://github.com/msracver/Deformable-ConvNets包括理解!Deformable Convolutional Networks摘要卷积神经网络(CNNs)由于其构建模块中固定的几何结构,其固有的局限性在于模型的几何变换。在这项工作中,我们引入了两个新的模块来增强...
2020-04-29 18:10:14
3700
原创 EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
参考文献:https://arxiv.org/pdf/1905.11946.pdf代码实现:https://github.com/lukemelas/EfficientNet-PyTorch包括理解!EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks摘要卷积神经网络(ConvNets)通常是在固...
2020-04-28 18:41:22
1010
原创 行人属性识别:HydraPlus-Net: Attentive Deep Features for Pedestrian Analysis
参考文献:https://arxiv.org/abs/1709.09930代码实现:https://github.com/xh-liu/HydraPlus-Net包括理解!HydraPlus-Net: Attentive Deep Features for Pedestrian Analysis摘要行人分析在智能视频监控中起着至关重要的作用,是以安全为中心的计算机视觉系统的关键组成部分。...
2020-04-14 10:21:24
3389
8
原创 行人属性识别:A Temporal Attentive Approach for Video-Based Pedestrian Attribute Recognition
参考文献:https://arxiv.org/abs/1901.05742代码实现:https://github.com/yuange250/video_pedestrian_attributes_recognition包括理解!A Temporal Attentive Approach for Video-Based Pedestrian Attribute Recognition摘要...
2020-04-10 18:07:17
2295
原创 行人属性识别:Improving Pedestrian Attribute Recognition With Weakly-Supervised Multi-Scale Attribute……
参考文献:Tang CF, Sheng L, Zhang ZX, Hu XL. Improving Pedestrian Attribute Recognition With Weakly-Supervised Multi-Scale Attribute-Specific Localization[J].ICCV-2019.代码实现:https://github.com/chufengt/icc...
2020-04-09 18:44:12
9585
5
原创 行人属性识别:Grouping Attribute Recognition for Pedestrian with Joint Recurrent Learning
参考文献:Zhao X, Sang LF, Ding GG, Guo YC, Jin XM. Grouping attribute recognition for pedestrian with joint recurrent learning[C]. Twenty-Seventh International Joint Conference on Artificial Intelligence ...
2020-04-08 18:49:16
2543
原创 目标检测:Faster R-CNN
详细内容于:https://blog.youkuaiyun.com/cdknight_happy/article/details/89713295
2020-01-21 17:59:50
815
原创 目标检测:Fast R-CNN
详细内容于:https://blog.youkuaiyun.com/cdknight_happy/article/details/87925098
2020-01-21 17:59:30
837
原创 目标检测:R-CNN
详细内容于:https://blog.youkuaiyun.com/cdknight_happy/article/details/86551937
2020-01-21 17:59:15
345
原创 目标检测中多尺度:特征金字塔FPN_Feature Pyramid Networks for Object Detection
原始内容来源于:https://blog.youkuaiyun.com/cdknight_happy/article/details/100528127https://blog.youkuaiyun.com/WZZ18191171661/article/details/79494534包含理解!参考文献:https://arxiv.org/abs/1612.03144代码实现:http://www.yueye....
2020-01-21 17:58:25
4762
原创 目标检测中基于角点检测:CornerNet Detecting Objects as Paired Keypoints
原始内容来源于:https://blog.youkuaiyun.com/weixin_40414267/article/details/82379793参考文献:https://arxiv.org/abs/1808.01244pytorch代码实现:https://github.com/umich-vl/CornerNet包括理解!CornerNet: Detecting Objects as Pai...
2020-01-20 11:35:20
2412
原创 目标检测基于中心点:CenterNet Keypoint Triplets for Object Detectiontection
参考文献:https://arxiv.org/abs/1904.08189代码实现:https://github.com/Duankaiwen/CenterNet截至目前2019.04.19,CenterNet应该是one-stage目标检测方法中性能(精度)最好的方法!CenterNet: Keypoint Triplets for Object Detectiontection摘要在...
2020-01-17 12:02:24
4933
原创 目标检测:SSD Single Shot MultiBox Detector
内容来源于:https://blog.youkuaiyun.com/cdknight_happy/article/details/91994312https://blog.youkuaiyun.com/xiaohu2022/article/details/79833786https://blog.youkuaiyun.com/thisiszdy/article/details/89576389包括整理!论文:https:/...
2019-12-30 18:23:56
610
原创 目标检测:YOLOV3 An Incremental Improvement
原始内容来源于:https://blog.youkuaiyun.com/cdknight_happy/article/details/91793142https://www.jianshu.com/p/d13ae1055302https://blog.youkuaiyun.com/litt1e/article/details/88907542https://blog.youkuaiyun.com/leviopku/artic...
2019-12-30 11:42:50
888
原创 目标检测:YOLOV2 You Only Look Once
参考文献:Redmon J , Farhadi A . [IEEE 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) - Honolulu, HI (2017.7.21-2017.7.26)] 2017 IEEE Conference on Computer Vision and Pattern Recog...
2019-10-24 18:49:54
511
原创 目标检测:YOLO You Only Look Once
参考文献: Redmon J , Divvala S , Girshick R , et al. You Only Look Once: Unified, Real-Time Object Detection[J]. 2015.项目主页: http://pjreddie.com/darknet/yolo/You Only Look OnceRedmon J , Divvala S , Gir...
2019-10-24 18:18:16
1070
原创 车牌检测License Plate Detection and Recognition in Unconstrained Scenarios
自然场景中无约束车牌检测
2019-10-24 17:15:45
2651
4
原创 车牌字符识别LPRNet:License Plate Recognition via Deep Neural Networks
转载于:link https://blog.youkuaiyun.com/qq_37053885/article/details/82598916感谢博主的辛苦劳动!!LPRNet:License Plate Recognition via Deep Neural NetworksSergeyZherzdevex−Intel∗IOTGComputerVisionGroupsergeyzherzde...
2019-10-24 14:28:23
7264
10
原创 车牌字符识别中ctc loss损失函数理解
原始内容来源于:https://zhuanlan.zhihu.com/p/43534801感谢博主的辛苦劳动!!1 概述文字识别是图像领域的常见问题,针对自然场景图像中的文字识别,包括两个步骤:首先进行文字检测,定位图像中文字位置;然后进行文字识别,将图像中的文字区域转换为字符信息。在车牌识别问题中,已定位好车牌位置,因此不需要进行文字检测,只需文字识别。(文字检测内容可参考https://...
2019-10-24 14:25:40
11792
1
原创 RNN中LSTM和GRU基础
原始内容来源于:https://blog.youkuaiyun.com/gzj_1101/article/details/79376798感谢博主的辛苦劳动!吴恩达深度学习课程理解更清晰!1 Recurrent Neural Networks(RNN)人类并不是每时每刻都从一片空白的大脑开始他们的思考。在你阅读这篇文章时候,你都是基于自己已经拥有的对先前所见词的理解来推断当前词的真实含义。我们不会将所...
2019-10-24 14:23:54
503
原创 车牌检测STN:Spatial Transformer Networks
参考文献:Max Jaderberg, Karen Simonyan, Andrew Zisserman, Koray Kavukcuoglu. Spatial Transformer Networks, 2016.linkSpatial Transformer Networks空间变换网络Max Jaderberg, Karen Simonyan, Andrew Zisserman, Kor...
2019-10-24 14:17:50
2073
空空如也
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人