zhengpeng7 / glcnet Goto Github PK
View Code? Open in Web Editor NEW[arXiv'21] Global-Local Context Network for Person Search.
License: MIT License
[arXiv'21] Global-Local Context Network for Person Search.
License: MIT License
@ZhengPeng7 hi thanks for the wonderful work and the code base , can you please share the pretrained weight file on google drive or on one drive
Thanks in advance
Hello. Thank you for great work.
I was wondering if adding your GCL module improves the performance in PRW and CUHK-SYSU datasets also.
Does the performance only improve in MovieNet test dataset?
您好!
当我将self.psn_feat_labelledOnly 设置为True时,出现了RuntimeError: Boolean value of Tensor with more than one value is ambiguous 这个错误,使用的数据集为cuhk,这是怎么回事呢?
When I run demo.py using comman as follow:
python3 demo.py --cfg configs/prw.yaml --ckpt ckpts/prw_469.pth
A size mismatch error occurs as follow:
Traceback (most recent call last):
File "demo.py", line 87, in
main(args)
File "demo.py", line 58, in main
resume_from_ckpt(args.ckpt, model)
File "/home/wbn/GLCNet/utils/utils.py", line 417, in resume_from_ckpt
model.load_state_dict(my_state_dict, strict=False)
File "/home/wbn/.conda/envs/glcnet2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1671, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for SeqNet:
size mismatch for roi_heads.embedding_head.projectors.feat_res4.0.weight: copying a param with shape torch.Size([256, 1024]) from checkpoint, the shape in current model is torch.Size([128, 2048]).
size mismatch for roi_heads.embedding_head.projectors.feat_res4.0.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128]).
size mismatch for roi_heads.embedding_head.projectors.feat_res4.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128]).
size mismatch for roi_heads.embedding_head.projectors.feat_res4.1.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128]).
size mismatch for roi_heads.embedding_head.projectors.feat_res4.1.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128]).
size mismatch for roi_heads.embedding_head.projectors.feat_res4.1.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128]).
I’m sure my pytorch version is correct, I don't konw why it happend.
Could you help me ? Thanks.
Thank you for your excellent work!I have two questions about network details.
1.Scene Context:
对于一张图片的每一个人来说,Scene Context 是怎么区别呢。是不是每个人所对应Scene Context都是相同的?都是该图片resnet最后输出的特征,经过CE模块后变为一个2048维的向量。应该是这样,但是我还想找你确认一下。
For everyone in a picture, how does Scene Context make the difference? Is the Scene Context the same for everyone? These are the features of the last output of the image resnet. After passing through the CE module, it becomes a vector of 2048 dimensions.It should be so, but I still want to check with you.
2.Group Context
在 Group CE之前,128维的特征向量是怎么来的?
您是把所有正样本的特征变成一个128维的特征吗?假如图中有两个人,那么有两个ROI区域。每一个ROI就有一个256维的特征向量,您把两个256的特征向量变维一个128维的向量。如果是三个人的话,就把三个256维的向量变为一个128维的向量。还有具体您是怎么实现的,直接cat变为256xN维的特征,然后再做一个1X1的卷积,变为128通道的特征,是这样吗?
Before Group CE, how did the 128-dimensional eigenvectors come from?
Are you turning all the features of a positive sample into a 128-dimensional feature? If there are two people in the graph, then there are two ROI regions. Each ROI has a 256-dimensional eigenvector, and you turn two 256 eigenvectors into a 128-dimensional vector. If it were three people, three vectors of 256 dimensions would be changed into one vector of 128 dimensions. And how exactly did you achieve this, directly cat into a 256*N-dimensional feature, and then do a 1X1 convolution to become a 128-channel feature, is that right?
How to make code changes to detect multiclass classification. 30 classes problem.
Is it even possible?
http://movienet.site unable to access, are there any other access methods
What is the difference?
What should I run to get the results in the paper?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.