yokattame / spindlenet Goto Github PK

same as the title.

hi,
I want to know your environment version like python version, ubuntu version, cuda and cudnn version.
ths!

I can not find out the experiments directory

Dear @yokattame
i am a fans of person reid, and your paper is amazing.
but i found some of dataset in your paper are difficult to down from the net when i reproduct your code.
for example:
(1)the link of 3DPeS_ReId_Snap is invalid.
(2)the official version of cuhk02 is .7z and it is different to .tar.gz that your code processed.
(3)the SenseReID is not open to public.
so, could your open your dataset?
and thanks for your patience.

I am new to person re-id. I have read your paper and found it amazing.
What kind of visualization method did you use to get the pictures in your paper?

I am very interest in your method and reading your code, but I can't find the 'Python' type layer as:

layer {
name: "data/spindle"
type: "Python"
top: "data"
top: "label"
top: "headbox"
top: "bodybox"
top: "legbox"
top: "rarmbox"
top: "larmbox"
top: "rlegbox"
top: "llegbox"
python_param {
module: "roi_data_layer"
layer: "RoiDataLayer"
param_str: "{'source': 'external/exp/datalists/jstl_10/train_p.txt', 'root_folder': '', 'batch_size': 100, 'new_height': 96, 'new_width': 96, 'shuffle': True, 'mirror': True, 'mean_value': [103.939,116.779,123.68], 'region_num': 7, 'region_scale': True}"
}
}

Can you help me with it?

Dear all,
I have trained the base model successfully but when I try to train the train_head using train_head.sh I got an error, see the attached photo. I know that the problem with this line
GLOG_log_dir=${log} ${CAFFE_DIR}/build/tools/caffe train --solver=${solver} --gpu=0
--weights=${pretrained_model}
I have one gpu (GeForce GTX 1070) installed. I play around change it but unfortunately ,no result.Any advice will be highly appreciated.

Dear all,
As we train 7 region parts, it is basically assumed that the region_num is different in the different models. For example, the region_num in the head_trainval. Prototxt should be different from the body_trainval. Prototxt while it is really identical (7)

So, how do you specify the region you focus on while traing different human parts??
Any help is highly appreciated.

I1104 10:02:52.843950 15737 layer_factory.hpp:77] Creating layer data/head
*** Aborted at 1541296973 (unix time) try "date -d @1541296973" if you are using GNU date ***
PC: @ 0x7f4387e33193 std::_Hashtable<>::clear()
*** SIGSEGV (@0x9) received by PID 15737 (TID 0x7f440eb1a780) from PID 9; stack trace: ***
@ 0x7f440c3b14b0 (unknown)
@ 0x7f4387e33193 std::_Hashtable<>::clear()
@ 0x7f4387e22a38 google::protobuf::DescriptorPool::FindFileByName()
@ 0x7f4387df8858 (unknown)
@ 0x7f440ca0e9f0 PyEval_EvalFrameEx
@ 0x7f440cb4405c PyEval_EvalCodeEx
@ 0x7f440ca9a46d (unknown)
@ 0x7f440ca6d273 PyObject_Call
@ 0x7f440ca8db75 (unknown)
@ 0x7f440ca24173 (unknown)
@ 0x7f440ca6d273 PyObject_Call
@ 0x7f440ca0b35c PyEval_EvalFrameEx
@ 0x7f440cb4405c PyEval_EvalCodeEx
@ 0x7f440ca05da9 PyEval_EvalCode
@ 0x7f440caa7244 PyImport_ExecCodeModuleEx
@ 0x7f440caa7c1f (unknown)
@ 0x7f440caa9390 (unknown)
@ 0x7f440caa9658 (unknown)
@ 0x7f440caaa76b PyImport_ImportModuleLevel
@ 0x7f440ca148b8 (unknown)
@ 0x7f440ca6d273 PyObject_Call
@ 0x7f440cb43487 PyEval_CallObjectWithKeywords
@ 0x7f440ca097e6 PyEval_EvalFrameEx
@ 0x7f440cb4405c PyEval_EvalCodeEx
@ 0x7f440ca05da9 PyEval_EvalCode
@ 0x7f440caa7244 PyImport_ExecCodeModuleEx
@ 0x7f440caa7c1f (unknown)
@ 0x7f440caa9390 (unknown)
@ 0x7f440caaa69c PyImport_ImportModuleLevel
@ 0x7f440ca148b8 (unknown)
@ 0x7f440ca6d273 PyObject_Call
@ 0x7f440cb43487 PyEval_CallObjectWithKeywords
./scripts/train_head.sh: line 25: 15737 Segmentation fault (core dumped) GLOG_log_dir=${log} ${CAFFE_DIR}/build/tools/caffe train --solver=${solver} --gpu=0 --weights=${pretrained_model}

Thanks your code!! It is very good and I have run the result,now I want to see some result of images or demo,so I want to ask you if you write some demo to display ,if you have it ,I hope you can share it !Thank you!!

Since the paper doesn't mention this information, I want to know the specific map number for more comparison with other algorithms

when I run "./scripts/train_head.sh",
get error "No module named roi_data_layer".

ps:my caffe version:"https://github.com/rbgirshick/py-faster-rcnn"

Hii I am interested in person re-ID. I have a clarifications, can you please provide answer for the following question?
"In your paper, for PRID datset, [17] perform better than SpindleNet but why in viper dataset, they didn't perform better than you?

It seems that "roi_data_layer.py" is missing for the training process of seven regions.
Could you provide that?

For example, for head region:
'python_param {
module: "roi_data_layer"
layer: "RoiDataLayer"
param_str: ...
}'

I am running the code only for Prid Dataset. In the "merge_datalists.sh" it needs file $DATALISTS/jstl_10". However no such directory is made when I run the previous 2 commands as mentioned in the README.
Also, what does this file/directory do?

Hi, we are trying to reproduce your learning system "spindle net" to
compare our research, but we hit the error while doing test.sh.
The learning processes by the scripts below looks successfully done.

./scripts/train_base.sh
./scripts/train_head.sh
./scripts/train_body.sh
./scripts/train_leg.sh
./scripts/train_rarm.sh
./scripts/train_larm.sh
./scripts/train_rleg.sh
./scripts/train_lleg.sh
./scripts/train_spindlenet.sh

But the "test.sh", I believe this is testing the result, died with the error.
Do you have any idea what cause this? or where should it be checked what happen?

Here is a log:

Extracting test_probe set
E0201 11:35:23.885982 10779 extract_features.cpp:52] Using GPU
E0201 11:35:23.886492 10779 extract_features.cpp:58] Using Device_id=0
processing 0
A total of 945 images.
terminate called after throwing an instance of 'boost::python::error_already_set'
Aborted (core dumped)

=our environment=
ubuntu16.04 + cuda8, (nvidia/cuda:8.0-cudnn6-devel-ubuntu16.04)

yokattame revision:
commit 1e88352569c4f4760be9b4509ccfed271c4805b6
Author: yokattame [email protected]
Date: Fri Jul 7 23:17:17 2017 +0800

caffe-master revision:
commit e963ef4eff86f07b67a23dfffc57f7e918959bd8
Author: Przemysaw Dolata [email protected]
Date: Wed Jan 3 10:21:03 2018 +0100

protobuf3.5.1.1-0: downloaded deb images from below
https://launchpad.net/~maarten-fonville/+archive/ubuntu/protobuf

Thanks in advance.

When I run ./scripts/test.sh, it shows error "terminate called after throwing an instance of 'boost::python::error_already_set", it seems the problem with the initialization with the first layer, but I do not have any idea about the specific problem. Do you know what is the wrong with it?

if the winners are always the max one？I’m very confused about it，would u give me some help sos，，，，，