wangguanan / aligngan Goto Github PK

Is there a visual process for searching?

Thank you. Is there a code for image retrieval in this code?

Time: 2019-12-04 13:19:29, indoor,multi, [array([0.56480978, 0.70538949, 0.77939312, 0.82884964, 0.86299819,
0.88858696, 0.90806159, 0.92327899, 0.93374094, 0.9423913 ,
0.94981884, 0.95425725, 0.95819746, 0.96259058, 0.96576087,
0.96893116, 0.97201087, 0.97423007, 0.97653986, 0.97817029,
0.98025362, 0.98201993, 0.98374094, 0.98532609, 0.98682065,
0.98817935, 0.98976449, 0.9910779 , 0.99221014, 0.99311594,
0.99397645, 0.99451993, 0.9951087 , 0.99560688, 0.99624094,
0.99660326, 0.99701087, 0.99737319, 0.99769022, 0.99787138,
0.99809783, 0.9982337 , 0.99850543, 0.9986413 , 0.99877717,
0.99877717, 0.99886775, 0.99895833, 0.9990942 , 0.99932065,
0.99950181, 0.9995471 , 0.99959239, 0.99977355, 0.99995471,
1. ]), 0.45644451754835486]

Thank for your work. I got this and the code stopped. Where are CMC and mAP values? I don't know if it's related to that I use two GPU with 32G memory and CPU with 24.0.

Hello, could you please share the The RegDB dataset?
My email: [email protected]
Thanks!

Can you detail the Baseline in the paper "RGB-Infrared Cross-Modality Person Re-Identification via Joint Pixel and Feature Alignment"? If you can give more about the Hyper-parameter settings， I would appreciate it.

The RegDB dataset can not be found in the official link ’http://dm.dongguk.edu/link.html‘
Could you please share this dataset？
Thanks！

Hello, when I downloaded the pre-training weight file (Baidu online disk), I found that the download extraction code zsr8 provided by you could not extract the file. I hope you can solve it. Thank you very much

There are seven models in the code:
self.model_list = [] self.model_list.append(self.G_rgb2ir) self.model_list.append(self.G_ir2rgb) self.model_list.append(self.D_rgb) self.model_list.append(self.D_ir) self.model_list.append(self.D_rgb_warmup) self.model_list.append(self.D_ir_warmup) self.model_list.append(self.encoder) self.model_list.append(self.embeder)
But it seems unclear in the paper. Can you explain the relationship bwteen them and the models in the paper, especially D_rgb and D_rgb_warmup?

The RegDB dataset can not be found in the official link ’http://dm.dongguk.edu/link.html‘
Could you please share this dataset？ [email protected]
Thanks！

Hello. I download the dataset, sysu-mm, from the provided baiduyun link. And when I run your code, I got the statistic of this dataset as:

And I got the performance result as:

which is inferior to that reported in the paper.

I wonder whether the problem of dataset causes inferior performance.
By the way, I use 4 1080Ti GPUs with 30G memory. My environment setting is: pytorch==1.1.0, python=3.7.6, torchvision=0.2.0.

Thanks a lot.

Thinks for your work. I want to ask why all the images need to convert to RGB includes IR images when loading them. I find it in this line of code: 'Image.open(img_path).convert('RGB')'.

hello, the link you provided is not work, so can you uplode the model to baidudisk? By the way, could you share the RegDB dataset with me?
email: [email protected]
Thanks

Thanks for your code.
Can I achieve the best result by trainging the model with less than 40G GPU memory(single GPU)?