abnerhqc/gaitset

A flexible, effective and fast cross-view gait recognition network

gaitset's Issues

Extracting human contours

Hello, if my dataset, the background in the video screen changes, is not stable, can I extract the human contour map very well? For example, there are other pedestrians and vehicles behind the pedestrians. I think it should have an impact, and the impact is still relatively large. That is to say, if there are other moving objects in the picture, the human contour cannot be extracted.

一个人最少要录几个视屏用于训练

我今天做测试的时候，一个人只是录了一个视屏得到一份轮廓图，但是我的文件夹里面只有一份数据轮廓数据就会报错，请问这个问题我该如何解决，虽然可以复制一份相同的得到两份数据，但是这样的话太占用空间了，能不能从代码上解决

这是两份数据

这是一份数据的情况，会报错，下标越界

如何预测

我现在用实际的数据进行训练了，我现在再用一组轮廓数据进行预测，我该如何做？

运行错误ModuleNotFoundError: No module named 'numpy.core._multiarray_umath'

作者你好，我在跑train.py时候在data_loader.py出现了ModuleNotFoundError: No module named 'numpy.core._multiarray_umath'，但是服务器上已经有numpy模块，网上查找的方法都不适合，请问需要如何解决这个错误？

CHANNEL ？

你好，打扰了，我现在预处理已经做好了，但是在我运行的时候，一直提示 return getattr(obj, method)(*args, **kwds)
AttributeError: 'NoneType' object has no attribute 'reshape' 关于reshape的问题。
我定位的错误应该是
frame_list = [np.reshape(
cv2.imread(osp.join(flie_path, _img_path)),
[self.resolution, self.resolution, -1])[:, :, 0]
for _img_path in imgs
if osp.isfile(osp.join(flie_path, _img_path))]
是这句话，这个是因为通道数的问题吗，我预处理后保存的图像是灰度图，但是cv2.imread读取后不是会自动转换为3通道吗，谢谢，期待你的回复

训练模型时报错： raise ValueError("Sample larger than population or is negative")

(dl) root@a4b3f83c3f3c:/GaitSet-master# python train.py
/usr/local/miniconda3/envs/dl/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 96, got 88
return f(*args, **kwds)
/usr/local/miniconda3/envs/dl/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 96, got 88
return f(*args, **kwds)
Initialzing...
Initializing data source...
Loading training data...
Data initialization complete.
Initializing model...
Model initialization complete.
Training START
Traceback (most recent call last):
File "train.py", line 21, in
m.fit()
File "/GaitSet-master/model/model.py", line 150, in fit
for seq, view, seq_type, label, batch_frame in train_loader:
File "/usr/local/miniconda3/envs/dl/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 501, in iter
return _DataLoaderIter(self)
File "/usr/local/miniconda3/envs/dl/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 297, in init
self._put_indices()
File "/usr/local/miniconda3/envs/dl/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 345, in _put_indices
indices = next(self.sample_iter, None)
File "/GaitSet-master/model/utils/sampler.py", line 15, in iter
self.batch_size[0])
File "/usr/local/miniconda3/envs/dl/lib/python3.6/random.py", line 320, in sample
raise ValueError("Sample larger than population or is negative")
ValueError: Sample larger than population or is negative

作者您好，请问这个报错怎么解决呢？

报错

/usr/bin/python3.5 /pycharmProject/GaitSet-master/train.py
Initialzing...
Initializing data source...
Data initialization complete.
Initializing model...
Model initialization complete.
Training START
Traceback (most recent call last):
File "/pycharmProject/GaitSet-master/train.py", line 21, in
m.fit()
File "/pycharmProject/GaitSet-master/model/model.py", line 150, in fit
for seq, view, seq_type, label, batch_frame in train_loader:
File "/home/tbb/.local/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 819, in iter
return _DataLoaderIter(self)
File "/home/tbb/.local/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 584, in init
self._put_indices()
File "/home/tbb/.local/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 646, in _put_indices
indices = next(self.sample_iter, None)
File "/pycharmProject/GaitSet-master/model/utils/sampler.py", line 19, in iter
_index = random.choices(
AttributeError: module 'random' has no attribute 'choices'

Unexpected Results for 24-subject training on CASIA-B.

Hi,
I recently ran your code on CASIA-B dataset to replicate the experiments. However, I had trouble in achieving the results reported in the paper. Below is the results I got:

Loading the model of iteration 80000...
Transforming...
Evaluating...
Evaluation complete. Cost: 1:12:31.519665
===Rank-1 (Include identical-view cases)===
NM: 10.178,     BG: 5.941,      CL: 3.517
===Rank-1 (Exclude identical-view cases)===
NM: 5.810,      BG: 4.069,      CL: 2.557
===Rank-1 of each angle (Exclude identical-view cases)===
NM: [6.16 5.61 7.12 4.85 6.11 6.92 5.96 5.56 6.42 5.26 3.94]
BG: [3.79 4.24 4.60 3.67 5.00 4.85 3.94 3.45 3.60 3.64 3.96]
CL: [1.62 2.78 2.48 3.20 3.03 3.24 2.98 2.49 2.49 2.26 1.57]

The training outputs are like these:

iter 74000:, hard_loss_metric=0.20000003, full_loss_metric=0.20000003, full_loss_num=229376.00000000, mean_dist=0.00000000, lr=0.000100, hard or full='hard'
0:03:51.626398
iter 75000:, hard_loss_metric=0.20000003, full_loss_metric=0.20000003, full_loss_num=229376.00000000, mean_dist=0.00000000, lr=0.000100, hard or full='hard'
0:03:51.586907
iter 76000:, hard_loss_metric=0.20000003, full_loss_metric=0.20000003, full_loss_num=229376.00000000, mean_dist=0.00000000, lr=0.000100, hard or full='hard'
0:03:51.859781
iter 77000:, hard_loss_metric=0.20000003, full_loss_metric=0.20000003, full_loss_num=229376.00000000, mean_dist=0.00000000, lr=0.000100, hard or full='hard'
0:03:46.837857
iter 78000:, hard_loss_metric=0.20000003, full_loss_metric=0.20000003, full_loss_num=229376.00000000, mean_dist=0.00000000, lr=0.000100, hard or full='hard'
0:03:46.709837
iter 79000:, hard_loss_metric=0.20000003, full_loss_metric=0.20000003, full_loss_num=229376.00000000, mean_dist=0.00000000, lr=0.000100, hard or full='hard'
0:03:49.217351
iter 80000:, hard_loss_metric=0.20000003, full_loss_metric=0.20000003, full_loss_num=229376.00000000, mean_dist=0.00000000, lr=0.000100, hard or full='hard'
Training COMPLETE

The triplet losses were 0.20000003 from about 10000th iteration and kept same till the end of training.
I have also tried to stop at 10000th iteration, but I got similar results as above.
Due to computational resources, I have not ran the 62-subject and 73-subject experiments.

I have ran the pretreatment.py following the instruction, and it worked well, only gave some warnings.
I have not modified the config.py, except for the 'pid_num' and paths.
I also modified the random.choices to np.random.choice, because the former one is a new feature in Python 3.6, while I used 3.5.
I have not modified other codes. Can you give any help? Thanks.

test.py value error

Initialzing...
Initializing data source...
Data initialization complete.
Initializing model...
Model initialization complete.
Loading the model of iteration 10000...
Transforming...
Traceback (most recent call last):
File "test.py", line 42, in
test = m.transform('test', opt.batch_size)
File "/media/cyprus/Data/DL/GaitSet-master/model/model.py", line 252, in transform
return np.concatenate(feature_list, 0), view_list, seq_type_list, label_list
ValueError: need at least one array to concatenate

作者大大您好，我迭代设置的10000次，sid_num30，出现这个报错的原因是什么呢？

请教下识别单个行人轮廓序列的做法

请问下如果想要实现只识别某一个特定角度例如角度为90°的行人轮廓序列，在训练方面，只有单个角度和单个行走状态的25张以上样本序列，足够用来训练吗？（在训练好项目中73个人模型的前提下，进行微调）。如果可以，在训练完后，在识别分类方面，有25张以上的行人轮廓序列，应该修改哪些代码或者应该以什么方式实现单人识别。希望可以给一些建议，谢谢您~

-

训练好的模型怎么样简单用于提取步态特征？

在test中，只是测试了数据集中的测试集，给出了结果。请问程序中有没有哪个地方可以得到提取步态特征的接口？或者说从哪里着手比较简单？
另外，如何加载一个训练好的模型同时不加载任何数据集（我看到test中是需要数据集的）？而采用在后续接口中输入数据的方式，我觉得这样更符合使用习惯。
因为对代码结构不太了解，对pytorch也是刚入门，以上问题想请教您该怎么入手解决？

代码问题

您好，这两个代码没看懂
我现在只用nm-01h和nm-02且只有90度角的数据进行训练和测试，请问下面这两个代码该如何修改

probe_seq_dict = {'CASIA': [['nm-05', 'nm-06'], ['bg-01', 'bg-02'], ['cl-01', 'cl-02']],
'OUMVLP': [['00']]}
gallery_seq_dict = {'CASIA': [['nm-01', 'nm-02', 'nm-03', 'nm-04']],
'OUMVLP': [['01']]}

在Test.py时遇到了value Error 求作者大大帮忙..

您好， AbnerHqc！这是我第二次发帖了（上次还在问能不能问windows10做 然后这段时间换成了ubuntu）感谢你在百忙之中抽空！我在运行test.py文件时候 遇到了一个bug：
Initialzing...
Initializing data source...
Data initialization complete.
Initializing model...
Model initialization complete.
Loading the model of iteration 80000...
Transforming...
Evaluating...
Traceback (most recent call last):
File "test.py", line 44, in
acc = evaluation(test, conf['data'])
File "/home/chenwei/下载/GaitSet-master/model/utils/evaluator.py", line 47, in evaluation
0) * 100 / dist.shape[0], 2)

！！！！！ValueError: could not broadcast input array from shape (0) into shape (5)

我在网络上得知： 报这种类型的错误，因为你输入数据的尺寸和网络需要的尺寸不一致，才会报错
所以很想请教下您，请问在运行test.py之前需要改什么吗 我除了config.py文件改了一些参数，其他都没有改。
（PS：我早在运行train.py训练模型的时候，config.py中 已经把dataset_path改成了 预处理之后的数据集B 的路径 ；其次我的pid_num也改成了30；然后，我的config.py中的batch_size改成了（2，4）因为之前我上面都没改动的时候也会发生Memory Error 提示我内存不够 当时我没改cache的值 ）
难道是因为batch_size的原因吗？
诚挚的希望你能够帮助我 谢谢！

图片裁剪

OU-MVLP identical-view

你好，AbnerHqC
非常感谢你提供开源代码，我在CASIA-B数据集上，成功复现了README中的rank@1。但是，在CASIA-B数据集上面训练的模型，在OU-MVLP上的rank@1很低，rank1=45.01%。
我想咨询下，在README中记录的OU-MVLP数据集上 Rank@1=87.1%，是使用CASIA-B数据集训练的还是OU-MVLP数据集训练的？

训练集最少个数

我今天想看看训练一个人要多少时间，把pid_num写成1，然后报错了

Traceback (most recent call last):
File "/home/zhanghua/PycharmProjects/Convolutional-Neural-Network-Based-Gait-Recognition-master/GaitSet-master/train.py", line 21, in
m.fit()
File "/home/zhanghua/PycharmProjects/Convolutional-Neural-Network-Based-Gait-Recognition-master/GaitSet-master/model/model.py", line 167, in fit
) = self.triplet_loss(triplet_feature, triplet_label)
File "/home/zhanghua/.local/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in call
result = self.forward(*input, **kwargs)
File "/home/zhanghua/.local/lib/python3.6/site-packages/torch/nn/parallel/data_parallel.py", line 143, in forward
outputs = self.parallel_apply(replicas, inputs, kwargs)
File "/home/zhanghua/.local/lib/python3.6/site-packages/torch/nn/parallel/data_parallel.py", line 153, in parallel_apply
return parallel_apply(replicas, inputs, kwargs, self.device_ids[:len(replicas)])
File "/home/zhanghua/.local/lib/python3.6/site-packages/torch/nn/parallel/parallel_apply.py", line 83, in parallel_apply
raise output
File "/home/zhanghua/.local/lib/python3.6/site-packages/torch/nn/parallel/parallel_apply.py", line 59, in _worker
output = module(*input, **kwargs)
File "/home/zhanghua/.local/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in call
result = self.forward(*input, **kwargs)
File "/home/zhanghua/PycharmProjects/Convolutional-Neural-Network-Based-Gait-Recognition-master/GaitSet-master/model/network/triplet.py", line 23, in forward
hard_hn_dist = torch.min(torch.masked_select(dist, hn_mask).view(n, m, -1), 2)[0]
RuntimeError: cannot perform reduction function min on tensor with no elements because the operation does not have an identity

Process finished with exit code 1

关于OUMVLP训练时的显存问题？

您好，我现在在OUMVLP数据集上做训练，但是在使用文章中参数时，batch_size为(32, 16)，(C1=C2=64; C3=C4=128; C5=C6=256)，运行报错，说是显存不足。但是我使用了4块16G显卡，请问下，您在做实验时使用了多少块显卡？

关于sp

您好，如果我想要修改sp相关内容，应该注意哪些代码呢？十分感谢您的帮助

treatise

Is there a treatise? I want to learn.

cannot reshape array of size 230400 into shape (64,64,newaxis)

How this issue to solve ? Can you help me ?

-

应用推广限制

你好，AbnerHqC
我在CASIA-B以及OU数据集上，复现了你们的成果，取得了很好的行人识别率。但是，在实际场景中，我们很难取得比较号的行人轮廓图片。
不知道你们有没有考虑，从RGB彩色图片中获取步态信息，而不是从黑白轮廓图中获取？

set-level feature and set-level feature

您好，请问set-level feature 和set-level feature应该怎么理解呢？？十分需要您的帮助。谢谢

creating our own test data(silhouette)

@AbnerHqC Thanks for sharing your excellent work. If I want to test GaitSet on my own data, how should I create the silhouette? I was wondering how to input my own data to the network and test the accuracies.
If I want to input the video, what should be the approach? (is it the same as creating the silhouettes from the video and using it as an input)
Sorry if my questions are too basic. Thank you.

参数 "batch_frame"实际意义？

代码gaitset.py中参数batch_frame的实际意义是什么？阅读代码的过程中有些困惑，还望解答。
另外，想确认下其他参数的实际含义，理解有偏差的地方还请指正：
n: batch_size default:(p,k) p:person_num k:sample_num;
s: frame_num default:30

Where does "DataParallelWithCallback" in GaitSet/model/model.py come from?

The following line seems strange in 'model.py'?
from sync_batchnorm import DataParallelWithCallback
Should there exist a file named 'sync_batchnorm.py‘ in 'model' directory?

execute pretreatment error!!!

first: I unzip datasetb and put dataset at dir '/root/DatasetB/silhouettes' like folder 001 002 003 and so on,
second: I run python3 pretreatment.py --input_path='/root/DatasetB/silhouettes/' --output_path='/root/DatasetB/output/',then i got the following warning. And no data generated !
JOB 440 : --WARNING-- seq:005-bg-01-000, less than 5 valid data.
JOB 448 : --WARNING-- seq:005-bg-01-144, frame:005-bg-01-144-061.png, no data, 0.
JOB 448 : --WARNING-- seq:005-bg-01-144, frame:005-bg-01-144-062.png, no data, 0.
JOB 448 : --WARNING-- seq:005-bg-01-144, frame:005-bg-01-144-063.png, no data, 0.
JOB 448 : --WARNING-- seq:005-bg-01-144, frame:005-bg-01-144-064.png, no data, 0.
JOB 448 : --WARNING-- seq:005-bg-01-144, frame:005-bg-01-144-065.png, no data, 0.
JOB 449 : --WARNING-- seq:005-bg-01-162, frame:005-bg-01-162-001.png, no data, 0.
JOB 449 : --WARNING-- seq:005-bg-01-162, frame:005-bg-01-162-002.png, no data, 0.

SyntaxError: invalid syntax

how to get out from this ?

-

-

您好，请教您几个问题

1、在运行train.py时，报下面错误：
Initialzing...
Initializing data source...
Loading training data...
Data initialization complete.
Initializing model...
Model initialization complete.
Training START
Traceback (most recent call last):
File "train.py", line 21, in
m.fit()
File "/home/xkk/Desktop/GaiSet/GaitSet-master/model/model.py", line 150, in fit
for seq, view, seq_type, label, batch_frame in train_loader:
File "/home/software_mount/xkk/local_install/python3/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 451, in iter
return _DataLoaderIter(self)
File "/home/software_mount/xkk/local_install/python3/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 247, in init
self._put_indices()
File "/home/software_mount/xkk/local_install/python3/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 295, in _put_indices
indices = next(self.sample_iter, None)
File "/home/xkk/Desktop/GaiSet/GaitSet-master/model/utils/sampler.py", line 15, in iter
self.batch_size[0])
File "/home/software_mount/xkk/local_install/python3/lib/python3.6/random.py", line 318, in sample
raise ValueError("Sample larger than population or is negative")
ValueError: Sample larger than population or is negative
2、在运行test.py时，报下面错误：
Initialzing...
Initializing data source...
Data initialization complete.
Initializing model...
Model initialization complete.
Loading the model of iteration 80000...
Transforming...
Traceback (most recent call last):
File "test.py", line 42, in
test = m.transform('test', opt.batch_size)
File "/home/xkk/Desktop/GaiSet/GaitSet-master/model/model.py", line 259, in transform
return np.concatenate(feature_list, 0), view_list, seq_type_list, label_list
ValueError: need at least one array to concatenate

请问：
您知道是什么原因吗？可能问题有点简单，但调试了两天不知道哪里有问题，期待您的解答～

还有一个问题是，config.py中的'dataset_path'是填写预处理之后的图片路径吗？如果不是的话，那预处理之后的图片在什么时候使用，谢谢您

it occurs the errors that"Sample larger than population or is negative",how do i solver it?

-

What's the purpose of the resolution parameter?

I found that in data_set.py you tend to cut the image according to resolution setting.

self.img2xarray( path)[:, :, self.cut_padding:-self.cut_padding].astype( 'float32') / 255.0

Why should we use this operation to cut the raw 64x64 image to 44*64 one?

pid_list = np.load(pid_fname)，TypeError: _reconstruct: First argument must be a sub-type of ndarray

emmmm.....今天可能是动了什么包，导致了如下错误，望您解答一下！
Initialzing...
Initializing data source...
Traceback (most recent call last):
File "train.py", line 18, in
m = initialization(conf, train=opt.cache)[0]
File "/home/home_data/lz/GaitSet/model/initialization.py", line 57, in initialization
train_source, test_source = initialize_data(config, train, test)
File "/home/home_data/lz/GaitSet/model/initialization.py", line 15, in initialize_data
train_source, test_source = load_data(**config['data'], cache=(train or test))
File "/home/home_data/lz/GaitSet/model/utils/data_loader.py", line 42, in load_data
pid_list = np.load(pid_fname)
File "/home/cv_user/anaconda3/lib/python3.6/site-packages/numpy/lib/npyio.py", line 433, in load
pickle_kwargs=pickle_kwargs)
File "/home/cv_user/anaconda3/lib/python3.6/site-packages/numpy/lib/format.py", line 657, in read_array
array = pickle.load(fp, **pickle_kwargs)
TypeError: _reconstruct: First argument must be a sub-type of ndarray

Process Problem

你好AbnerHqC，使用CASIA-B数据集，未做任何修改。在加载数据阶段出现以下问题

经过计算，原始图片是 320 * 240 * 3 的 ndrray。 reshape到64*64 剩余部分会出现小数导致错误。不知道你那边是如何处理的？

training START

在执行train.py后 报以下错误 memory error请问这如何解决？感谢您。我的配置是GTX950 cuda9 csdnn7 py3.6 pytorch0.4

Initialzing...
Initializing data source...
Loading training data...
Data initialization complete.
Initializing model...
Model initialization complete.
Training START
Traceback (most recent call last):
File "train.py", line 21, in
m.fit()
File "D:\DL\GaitSet-master\model\model.py", line 150, in fit
for seq, view, seq_type, label, batch_frame in train_loader:
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\site-packages\torch\utils\data\dataloader.py", line 501, in iter
return _DataLoaderIter(self)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\site-packages\torch\utils\data\dataloader.py", line 289, in init
w.start()
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\multiprocessing\process.py", line 105, in start
self._popen = self._Popen(self)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\multiprocessing\context.py", line 223, in _Popen
return _default_context.get_context().Process._Popen(process_obj)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\multiprocessing\context.py", line 322, in _Popen
return Popen(process_obj)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\multiprocessing\popen_spawn_win32.py", line 65, in init
reduction.dump(process_obj, to_child)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\multiprocessing\reduction.py", line 60, in dump
ForkingPickler(file, protocol).dump(obj)
MemoryError
Initialzing...
Initializing data source...
Loading training data...
Traceback (most recent call last):
File "", line 1, in
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\multiprocessing\spawn.py", line 105, in spawn_main
exitcode = _main(fd)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\multiprocessing\spawn.py", line 114, in _main
prepare(preparation_data)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\multiprocessing\spawn.py", line 225, in prepare
_fixup_main_from_path(data['init_main_from_path'])
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\multiprocessing\spawn.py", line 277, in _fixup_main_from_path
run_name="mp_main")
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\runpy.py", line 263, in run_path
pkg_name=pkg_name, script_name=fname)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\runpy.py", line 96, in _run_module_code
mod_name, mod_spec, pkg_name, script_name)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\runpy.py", line 85, in _run_code
exec(code, run_globals)
File "D:\DL\GaitSet-master\train.py", line 18, in
m = initialization(conf, train=opt.cache)[0]
File "D:\DL\GaitSet-master\model\initialization.py", line 57, in initialization
train_source, test_source = initialize_data(config, train, test)
File "D:\DL\GaitSet-master\model\initialization.py", line 18, in initialize_data
train_source.load_all_data()
File "D:\DL\GaitSet-master\model\utils\data_set.py", line 45, in load_all_data
self.load_data(i)
File "D:\DL\GaitSet-master\model\utils\data_set.py", line 48, in load_data
return self.getitem(index)
File "D:\DL\GaitSet-master\model\utils\data_set.py", line 62, in getitem
data = [self.loader(_path) for _path in self.seq_dir[index]]
File "D:\DL\GaitSet-master\model\utils\data_set.py", line 62, in
data = [self.loader(_path) for _path in self.seq_dir[index]]
File "D:\DL\GaitSet-master\model\utils\data_set.py", line 53, in loader
'float32') / 255.0
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\site-packages\xarray\core\dataarray.py", line 1954, in func
**kwargs))
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\site-packages\xarray\core\ops.py", line 207, in func
return _call_possibly_missing_method(self, name, args, kwargs)
File "C:\Users\Cyprus\AppData\Local\Programs\Python\Python36\lib\site-packages\xarray\core\ops.py", line 194, in _call_possibly_missing_method
return method(*args, **kwargs)
MemoryError

What exactly the dataset looks like when running test.py?

def __loader__(self, path):
        file_path = osp.join(path, os.listdir(path)[0])
        seq = pickle.load(open(file_path, 'rb'))
        data_type = seq.attrs['data_type']
        data_name = seq.name
        if data_name in self.selector:
            seq = seq.loc.__getitem__(tuple(self.selector[data_name]))
        if data_type == 'img':
            seq = seq.astype('float32')
            seq /= 255.0
            seq = seq[:, :, 10:54]
        return seq

in data_set.py, what the pickle tries to load, not a image?

Prepocess

你好，AbnerHqcz！
我想咨询下，这个训练文件夹GaitDataSet中s的图片是不是都要归一化到88*128的尺寸呢？并且归一化的方法中提高了滑动均值滤波，这个滑动帧数是30帧吗？那最后的三十帧怎么滑动呢？是循环滑动均值吗？

谢谢，期待你的回复

Happy New Year,Abnerhqc！Can these codes work on windows10 instead of linux？

sincerely hope you can help me thanks!

INPUT_PATH

I fill (INPUT_PATH) with this :

line 44: opt.input_path = ('D:\Road.to.Sidang\Python\GaitRecog\Master\gei01\001\bg-01')

....

line 164: seq_type = os.listdir(os.path.join("d:", os.sep, "Road.to.Sidang", "Python", "GaitRecog", "Master", "gei01", "001", "bg-01"), _id))

nb: my input datarset directory is D:\Road.to.Sidang\Python\GaitRecog\Master\gei01

would you like to help me?

live camera test

I want to know how to test on camera if very frame have many people

After I was training the whole CASIA-B dataset.

about align datasets methods

could you release your alignment method code?

get value bigger than 100 when test OUMVLP dataset

Hi,
I modified the code to train model on OUMVLP dataset. And when I try to test it, I find that the accuracy rate is bigger than 100 when Exclude identical-views. I wonder whether this is because lack of some views in test set?

===Rank-1 (Include identical-view cases)===
NM: 80.799
===Rank-1 (Exclude identical-view cases)===
NM: 104.309
===Rank-1 of each angle (Exclude identical-view cases)===
NM: [ 90.00 110.00 96.67 110.00 112.00 111.67 108.33 115.00 101.67 98.33
101.67 101.67 101.67 101.67]

Can you help me about this?
Thanks so much

FileNotFoundError: [WinError 2] The system cannot find the file specified: './work' in train.py

-

请问下有比较好的项目用来提取视频轮廓图吗？

你好，请问下对于你们的项目，有什么推荐的方式用来通过opencv视频流提取轮廓图用来训练吗?目前用尝试用Vibe+进行视频轮廓图提取，但是效果不大好，提取的速度太慢以至于没多大的实用性。DensePose这项目怎么样，可以给点建议吗？谢谢你~

how to understand the distance method in triplet.py

Hi,
I am just a little confused about the distance method in the triplet.py.
During training, the input x size is (batchsize, 62, hidden_dim). So it should have batchsize features for each single image set and we want the distance between them.
the code below calculate the distance for batch all. And I have difficult when I try to understand how this can work.
x2 = torch.sum(x ** 2, 2)
dist = x2.unsqueeze(2) + x2.unsqueeze(2).transpose(1, 2) - 2 * torch.matmul(x, x.transpose(1, 2))
dist = torch.sqrt(F.relu(dist))
I would appreciate if you guys can help me understand this code or give me some reference to read.
Thanks so much.