cleardusk / emotionchallenge

Source code for 1st winner of face micro-emotion competition, FG 2017.

Hi there, I'm Jianzhu Guo 👋

emotionchallenge's Issues

SyntaxError: invalid syntax:: caffe train -solver solver.prototxt -gpu 3 2>&1 | tee final_03051346.log

So, I have a couple of questions and I really appreciate if you could help:

1. Do you have pretrained model in which I could simply feed in an image and get the emotion out of your system? if that is the case where should I feed in the image?

2. How are you doing the training process? I don't get where is the past to the training dataset? It is not mentioned in your Read.me neither it is mentioned which dataset you are using for training. Can you please share those information?

Thanks a lot,
Mona

Here are all the steps I took but really don't make any sense because I don't even know how to feed in a test image or how to set the path to training images

[jalal@goku cs585]$ cd EmotionChallenge/
[jalal@goku EmotionChallenge]$ ls
cnn  crop_align  data  fact_sheets.pdf  fact_sheets.tex  models  readme.md
[jalal@goku data]$ file train_data
train_data: broken symbolic link to `/data/gjz/Training'
[jalal@goku data]$ cd ..
[jalal@goku EmotionChallenge]$ cd crop_align/
[jalal@goku crop_align]$ mkdir build
[jalal@goku crop_align]$ cd build/
[jalal@goku build]$ cmake ..
-- The C compiler identification is GNU 4.8.5
-- The CXX compiler identification is GNU 4.8.5
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- OpenCV:         /scratch/sjn-p2/anaconda/anaconda2/include/opencv/scratch/sjn-p2/anaconda/anaconda2/include
-- Configuring done
-- Generating done
-- Build files have been written to: /scratch/mona/download/cs585/EmotionChallenge/crop_align/build
[jalal@goku build]$ make
Scanning dependencies of target crop_align
[ 33%] Building CXX object CMakeFiles/crop_align.dir/crop_align.cpp.o
[ 66%] Building CXX object CMakeFiles/crop_align.dir/imtransform.cpp.o
[100%] Linking CXX executable crop_align
[100%] Built target crop_align
[jalal@goku build]$ cd ..
[jalal@goku crop_align]$ ls
build  CMakeLists.txt  crop_align.cpp  crop_align.py  imtransform.cpp  imtransform.h  __init__.py  landmark.py  util.py
[jalal@goku crop_align]$ cd ..
[jalal@goku EmotionChallenge]$ cd cnn
[jalal@goku cnn]$ ls
data_layer.py  deploy.prototxt  extract.py  landmark_224.py  net.prototxt  prepare_data.py  solver.prototxt  train_val.sh  util.py
[jalal@goku cnn]$ python prepare_data.py 
[jalal@goku cnn]$ python train_val.sh 
  File "train_val.sh", line 3
    caffe train -solver solver.prototxt -gpu 3 2>&1 | tee final_03051346.log
              ^
SyntaxError: invalid syntax

Additionally, I am not sure what the line of code that is causing error is trying to do and what is the fix to it.

caffe: symbol lookup error: caffe: undefined symbol: _ZN6google14FlagRegistererC1EPKcS2_S2_S2_PvS3_

Additionally, after I run

 1168  cd cnn/
 1169  ./prepare_data.py
 1192  ./train_val.sh

I get this error:
[jalal@goku cnn]$ ./train_val.sh caffe: symbol lookup error: caffe: undefined symbol: _ZN6google14FlagRegistererC1EPKcS2_S2_S2_PvS3_

Do you how should this be fixed?

TypeError: object of type 'map' has no len()

Using Python2.7 I get this error:

(py2emotion) [jalal@goku crop_align]$ /scratch/sjn-p2/anaconda/anaconda2/bin/python landmark.py
Traceback (most recent call last):
  File "landmark.py", line 7, in <module>
    import dlib
  File "/scratch/sjn-p2/anaconda/anaconda2/lib/python2.7/site-packages/dlib/__init__.py", line 1, in <module>
    from .dlib import *
ImportError: libcudnn.so.6: cannot open shared object file: No such file or directory
(py2emotion) [jalal@goku crop_align]$ which conda
/scratch/anaconda3/envs/py2emotion/bin/conda
(py2emotion) [jalal@goku crop_align]$ source deactivate py2emotion
[jalal@goku crop_align]$ /scratch/sjn-p2/anaconda/anaconda2/bin/python landmark.py
Traceback (most recent call last):
  File "landmark.py", line 7, in <module>
    import dlib
  File "/scratch/sjn-p2/anaconda/anaconda2/lib/python2.7/site-packages/dlib/__init__.py", line 1, in <module>
    from .dlib import *
ImportError: libcudnn.so.6: cannot open shared object file: No such file or directory
[jalal@goku crop_align]$ /scratch/sjn-p2/anaconda/anaconda2/bin/conda install cudnn=6.0
Fetching package metadata .............
Solving package specifications: .

UnsatisfiableError: The following specifications were found to be in conflict:
  - caffe-gpu -> cudnn ==5.1
  - cudnn 6.0*
Use "conda info <package>" to see the dependencies for each package.

[jalal@goku crop_align]$

Using Python 3, I get this other error. No error on python landmark.py but in the command following it:

[jalal@goku crop_align]$ python landmark.py
[jalal@goku crop_align]$ cd ..
[jalal@goku EmotionChallenge]$ ls
cnn  crop_align  data  fact_sheets.pdf  fact_sheets.tex  models  readme.md
[jalal@goku EmotionChallenge]$ cd cnn/
[jalal@goku cnn]$ ls
data_layer.py  deploy.prototxt  extract.py  landmark_224.py  net.prototxt  prepare_data.py  __pycache__  solver.prototxt  train_val.sh  util.py
[jalal@goku cnn]$ python extract.py
Traceback (most recent call last):
  File "extract.py", line 78, in <module>
    submit()
  File "extract.py", line 74, in submit
    extract('../models/final.caffemodel')
  File "extract.py", line 58, in extract
    assert len(landmark) == 136
TypeError: object of type 'map' has no len()

If you could help to solve preferrably the one that gives an error for Python 3 and if not the one for Python 2.7. Thanks.

FG2020 Compound Emotion Challenge participant

我是FG2020复合表情的参赛者，请问你有上一届比赛的论文模板吗？另外最终测试集的解压密码也没有公布，想请问一下上一届的比赛最终测试阶段是什么流程？

crop and align doesn't work

So, I get this error for Python 3:

[jalal@goku cnn]$ python extract.py
Traceback (most recent call last):
  File "extract.py", line 78, in <module>

▽
    submit()
  File "extract.py", line 74, in submit
    extract('../models/final.caffemodel')
  File "extract.py", line 61, in extract
    pred = classifier.predict(img, landmark)
  File "extract.py", line 32, in predict
    self.net.blobs['data'].data[0, ...] = image_array.transpose((2, 0, 1))

▽
AttributeError: 'NoneType' object has no attribute 'transpose'

Looking at the code:

def extract(model_file):
    classifier = Classifier(model_file=model_file,
                            deploy_file='deploy.prototxt')
    records = open('../data/test_ld.txt').read().strip().split('\n')
    # print(records[:1])
    predicts = []
    for rec in records:
        rec = rec.split()
        fp = rec[0]
        # print(fp)
        landmark = rec[1:]
        landmark = map(float, landmark)
        # print(landmark)
        assert len(list(landmark)) == 136

It is looking at this file:
../data/test_ld.txt

Looking at this file:
I have something like
../data/face_224/104/Sample_104_001.JPG
I expect to see a folder named 104 in the face_224 folder. However, there is no folder here and only a readme.md file which says 224x224 size image stores here.

[jalal@goku data]$ cd face_224/
[jalal@goku face_224]$ ls
readme.md

So I am confused what has gone wrong and how to fix it?
Any suggestion is really appreciated. I had no problem with the first step which was done in the crop_align folder:

[jalal@goku crop_align]$ python landmark.py
[jalal@goku crop_align]$

From what I understand the crop didn't work because you are stating that you expect the cropped folder to be in data/face_224 folder in which face_224 is the name of my image however, there's nothing in that folder.

Any help is really appreciated.