Adding python data augmentation layer raising an issue "ImportError: cannot import name _remove_dead

[Shima Caffe]

Hi,

I am doing segmentation. Thus, both image and label need to be shifted in the same direction and same shifting value. I am trying to add augmentation layer to do shifting in x and y direction. 

What I am doing wrong in this Python Layer code? 

Really at a loss here, any help is greatly appreciated.

'''Usage
layer {
    name: 'aug_layer'
    type: 'Python'
    bottom: 'data'
    bottom: 'label'
    top: 'data'
    top: 'label'
    python_param {
        module: 'augLayer'
        layer: 'AugmentationLayer'
    }
    include{
    phase: TRAIN
    }
}
'''
import numpy as np
import caffe
import random
#from augmentation_main import doShift_xy
import os,sys
import h5py
import matplotlib.pyplot as plt

#import unittest
#import tempfile
#import os
import six


import  scipy
import scipy.io as sio
import os,sys

from skimage import transform as tf
from skimage.transform import AffineTransform
from skimage.util import random_noise


TRAIN =0
TEST =1
class AugmentationLayer(caffe.Layer):
    def setup(self,bottom,top):
        #assert len(bottom)==2            #requires two layer.bottom 1:image( N x K x H x W) 2:label ( N x 1 x H x W)
        #assert bottom[0].data.ndim>= 3   #requires image data
        if len(bottom) != 2:
            raise Exception("Wrong number of bottom blobs (data, label)")
        if len(top)!=2:
            raise Exception("Wrong number of top blobs (data,label)")
        self.img=[]
        self.lbl=[]
        self.totalImgs=0
        #self.phase=TRAIN


    def reshape(self,bottom,top):
        #top[0].reshape(*bottom[0].data.shape)
        #top[1].reshape(*bottom[1].data.shape)
        pass

    def forward(self,bottom,top):
        if self.phase == TRAIN:

            image=bottom[0].data
            label=bottom[1].data

            self.totalImgs += len(label)

            for i in range(len(label)):     #len(label) is equal to batch size
                img=image[i].transpose(1, 2, 0)       # Image: change from (K x H x W) to (H x W x K)
                #if img.shape[2]==3:     #if it is an RGB three channel image
                img=img[:,:,(2,1,0)]                  # change from BGR to RGB

                lbl=label[i].transpose(1, 2, 0)                           # Label: shape (1 x H x W) 
                lbl=lbl.reshape(lbl.shape[:-1])

                im,lb=self.doShift_xy(img,lbl)

                im=im[:, :, (0,1,2)].transpose(2, 0, 1)  # Change the channel from (RGB to BGR) and change from (H x W x K) to (K x H x W)     

                lb=lb.reshape(lb.shape[0],lb.shape[1], 1)
                lb=lb.transpose(2, 0, 1)                 # Change from (H x W x K) to (K x H x W)
                print('successfully tested')

                top[0].data[i,...]=im
                top[1].data[i,...]=lb

        elif self.phase ==TEST:
            pass

        def doShift_xy(self,img_,lbl_):
            num_channel=img_.shape[2]
            x_trans=random.randrange(-10,10)
            y_trans=random.randrange(-10,10)
            for i in range(0,num_channel):     #apply on the all channels of an image
                tmp=tf.warp(img_[:,:,i], AffineTransform(translation=(x_trans,y_trans)))
                img_[:,:,i]=tmp

            lbl_=tf.warp(lbl_, AffineTransform(translation=(x_trans,y_trans)))

            return img_,lbl_       


    def backward(self,top,propagate_down,bottom):
        pass

This is the error I am getting and could solve it, 

I0207 02:45:33.556780 19447 net.cpp:84] Creating Layer label
I0207 02:45:33.556783 19447 net.cpp:380] label -> label
I0207 02:45:33.556872 19447 data_layer.cpp:45] output data size: 1,1,87,256
I0207 02:45:33.557867 19447 net.cpp:122] Setting up label
I0207 02:45:33.557878 19447 net.cpp:129] Top shape: 1 1 87 256 (22272)
I0207 02:45:33.557880 19447 net.cpp:137] Memory required for data: 356352
I0207 02:45:33.557883 19447 layer_factory.hpp:77] Creating layer aug_layer
Traceback (most recent call last):
  File "/home/ubuntu/caffe/python/augmentLayer.py", line 22, in <module>
    import numpy as np
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/site-packages/numpy/__init__.py", line 142, in <module>
    from . import add_newdocs
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/site-packages/numpy/add_newdocs.py", line 13, in <module>
    from numpy.lib import add_newdoc
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/site-packages/numpy/lib/__init__.py", line 8, in <module>
    from .type_check import *
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/site-packages/numpy/lib/type_check.py", line 11, in <module>
    import numpy.core.numeric as _nx
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/site-packages/numpy/core/__init__.py", line 74, in <module>
    from numpy.testing import _numpy_tester
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/site-packages/numpy/testing/__init__.py", line 10, in <module>
    from unittest import TestCase
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/unittest/__init__.py", line 64, in <module>
    from .main import TestProgram, main
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/unittest/main.py", line 7, in <module>
    from . import loader, runner
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/unittest/runner.py", line 7, in <module>
    from .signals import registerResult
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/unittest/signals.py", line 2, in <module>
    import weakref
  File "/home/ubuntu/anaconda2/envs/testcaffe/lib/python2.7/weakref.py", line 14, in <module>
    from _weakref import (
ImportError: cannot import name _remove_dead_weakref

some users referring to environment, however, I have even tried on a fresh anaconda2 environment. But it did not solve my problem. What do you suggest?