The directory， pw = '/home/zzd/re_ID/market1501/bounding_box_train_256/'，in prepare_data is what？

[qiulesun]

No description provided.

[layumi]

First we need to resize the images to 256x256 for the following processing (like crop 224x224).
The resized images will store into this dir.

[qiulesun]

I got the error,'无法打开要写入的文件 "E:\project\STN\Pedestrian_Alignment-master\dataset\bounding_box_train_256\0002_c1s1_000451_03.jpg"',when run the imwrite(im,url256) in prepare_data.

[qiulesun]

I find the reson. The dictiontory, bounding_box_train_256, is not created.

[TaihuLight]

Why resize the images to 256 * 256 instead of 224 * 224, and the input size of ResNet-50 is 224*224?

[TaihuLight]

The network can be visulized with the web tool http://ethereon.github.io/netscope/
And the model can be converted to caffemodel?
https://github.com/marvis/pytorch-caffe-darknet-convert

[layumi]

Hi @TaihuLight
I use a trick. Since ResNet is a fully convolutional network, using 256*256 image as input will output 4*4*2048. Then I sum the first and second dimension and get the 1*1*2048 vector.
I found this feature is a little better.

[TaihuLight]

Why the name of the image in market1501 is *.jpg.jpg as follows ? It is the wrong or the trick ?

[layumi]

Hi @TaihuLight
I am not sure what is wrong.
Are these images generated when preparing data?
Anyway, I think it is fine since matlab still can read them.

[TaihuLight]

It is only a bug for the market1501 dataset, and these images are the original images in the official dataset market1501. I think your team will fix it in future.

[daySunny]

How to solve this problem that it doesn't hava market-1501/bounding_box_train_256?

[layumi]

Hi @daySunny
Have you run the prepare_data.m? The code will generate this folder.
( Sometime you may need to mkdir market-1501/bounding_box_train_256 first. Matlab do not have the right to mkdir)

[daySunny]

Hi,thanks for solve my question. I try to run the prepare.m,but there still some questions. In the function randi ,s equal 0.So, I can't run sucessfully. Can you give me some suggestions,thanks again. ------------------ Original ------------------ From: "Zhedong Zheng"<[email protected]> Date: Fri, May 4, 2018 10:33 AM To: "layumi/Pedestrian_Alignment"<[email protected]>; Cc: "daySunny"<[email protected]>;"Mention"<[email protected]>; Subject: Re: [layumi/Pedestrian_Alignment] The directory， pw = '/home/zzd/re_ID/market1501/bounding_box_train_256/'，in prepare_data is what？ (#2) Hi @daySunny Have you run the prepare_data.m? The code will generate this folder. &mdash; You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.