
论文
文章平均质量分 95
breeze_blows
这个作者很懒,什么都没留下…
展开
-
Mining Inter-Video Proposal Relations for Video Object Detection
Mining Inter-Video Proposal Relations for Video Object Detection. ECCV2020的一篇视频目标检测文章论文链接:https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123660426.pdf补充材料:https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123660426-supp.pdf代码:https:/原创 2020-09-25 17:48:29 · 1696 阅读 · 1 评论 -
常用的神经网络结构
1.resnet论文链接:https://arxiv.org/pdf/1512.03385.pdf网络结构:原创 2020-02-22 10:02:09 · 1401 阅读 · 0 评论 -
Temporal Context Enhanced Feature Aggregation for Video Object Detection
论文链接:https://www.aaai.org/Papers/AAAI/2020GB/AAAI-HeF.1752.pdfAAAI2020的一篇文章原创 2020-06-26 14:29:22 · 2276 阅读 · 0 评论 -
Dual Attention Network for Scene Segmentation
论文链接:https://arxiv.org/pdf/1809.02983.pdfCVPR2019的一篇文章代码链接:https://github.com/junfu1115/DANet/主要思想:按照文中的说法就是we append two types of attention modules on top of dilated FCN, which model the semantic interdependencies in spatial and channel dimensions re.原创 2020-06-19 09:58:21 · 694 阅读 · 0 评论 -
PolarMask: Single Shot Instance Segmentation with Polar Representation
论文链接:https://arxiv.org/pdf/1909.13226.pdfCVPR2020oral代码链接:https://github.com/xieenze/PolarMask原创 2020-06-11 10:35:59 · 373 阅读 · 0 评论 -
YOLOv4: Optimal Speed and Accuracy of Object Detection
论文链接:https://arxiv.org/pdf/2004.10934.pdf代码链接:https://github.com/AlexeyAB/darknet主要思想:利用各种universal features 包括Weighted-Residual-Connections (WRC), Cross-Stage-Partial-connections (CSP), Cross mini-Batch Normalization (CmBN), Self-adversarial-training原创 2020-05-29 08:59:14 · 700 阅读 · 0 评论 -
目标分割(semantic segmentation, instance segmentation... )论文阅读
记录一下泛读过的一些实例分割的论文,持续更新。。。原创 2020-05-22 09:19:47 · 399 阅读 · 0 评论 -
Res2Net: A New Multi-scale Backbone Architecture
论文链接:https://arxiv.org/pdf/1904.01169.pdf2020 IEEE TPAMI 2020的一篇文章代码链接:https://github.com/Res2Net/Res2Net-PretrainedModels主要思想:主要是为了增强resnet的multi-scale的能力,以往获得the multi-scale features的方法往往都是一种l...原创 2020-05-02 16:00:50 · 2376 阅读 · 0 评论 -
Improving Convolutional Networks with Self-Calibrated Convolutions
论文链接:http://mftp.mmcheng.net/Papers/20cvprSCNet.pdfCVPR2020的一篇文章代码:https://github.com/backseason/SCNet主要思想:SCnet,提出了一个self-calibrated convolutions,其实感觉就是在channel上做attention,按照文中的说法Benefiting from...原创 2020-05-02 12:11:40 · 3780 阅读 · 1 评论 -
视频目标检测(video object detection)简单综述
1.Memory2.Attention RDN: ICCV2019 Relation Distillation Networks for Video Object Detection[paper]. faster rcnn为backbone,用multi-stage的形式将support frame的proposal来逐步增强reference frames的proposal特征,...原创 2020-04-13 14:16:55 · 32451 阅读 · 14 评论 -
Mask R-CNN
we propose a simple, quantization-free layer, called RoIAlign, that faithfully preserves exact spatial locations. Despite being a seemingly minor changeSecond, we found it essential to decouple mask ...原创 2020-03-29 16:30:12 · 428 阅读 · 0 评论 -
Deformable Convolutional Networks && Region Proposal by Guided Anchoring
https://zhuanlan.zhihu.com/p/58185157原创 2020-03-21 16:22:50 · 1626 阅读 · 0 评论 -
GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond
for different query positions, their attention maps are almost the same. the global context after training is actually independent of query position,there is no need to compute query-specific global c...原创 2020-03-15 17:08:38 · 767 阅读 · 0 评论 -
计算机视觉(CV)中的注意力Attention机制
记录看到的一些不错的attention的文章1.Squeeze-and-Excitation Networks该方法是winners of ILSVRC 2017 classification competition。主要思想就是下面这幅图。通道数为C’的feature在经过一系列卷积变换Ftr之后变成通道数为C的feature,之后便分为三个步骤,首先是Squeeze即Fsq,用...原创 2020-03-13 11:35:31 · 2914 阅读 · 0 评论 -
Non-local Neural Networks
CVPR2018的一篇文章论文:https://arxiv.org/pdf/1711.07971v3.pdf主要贡献:以往的CNN或者RCNN都是考虑的local的信息,虽然可以通过扩大感受野来增加不那么local的信息,但是却是有限的,non local按照文中的说法就是non-local operation computes the response at a position as ...原创 2020-03-07 16:45:35 · 974 阅读 · 0 评论 -
Relation Distillation Networks for Video Object Detection
ICCV2019的一篇视频目标检测的文章论文链接:https://arxiv.org/pdf/1908.09511v1.pdf据文中描述代码基于pytorch1.0所写,只是现在仍未开源。。。主要贡献:设计了RDN(Relation Distillation Networks)模块,其实就是用两个stage来用support frame的proposal来逐步增强reference f...原创 2020-03-07 10:44:49 · 2576 阅读 · 0 评论 -
Relation Networks for Object Detection
#求rois之间的坐标关系,找到位置信息rois=[300,4]nongt_dim = 300position_matrix = extract_position_matrix(rois, nongt_dim=nongt_dim) #position_matrix=[300,300,4]#extract_position_embedding方法实现论文中公式5的EG操作posi...原创 2020-03-06 09:39:33 · 981 阅读 · 1 评论 -
视频目标检测论文
记录一些略看的视频目标检测论文。1.Sequence Level Semantics Aggregation for Video Object Detection(ICCV 2019)论文链接:https://arxiv.org/abs/1907.06390v2主要贡献:以faster rcnn为backbone设计了一个selsa模块来提高在视频目标检测数据集即VID中的精度,训练集...原创 2020-02-28 15:28:18 · 6660 阅读 · 6 评论