Tensorflow1.0 实现Sptial Transformer Networks遇到的问题

2017-02-28 11:01:06 3.4k 阅读 AI 写同款 · GEO 优化AI 选题数据监测一键分发 ›

解决OpenAI API连接错误：HTTPSConnectionPool问题的实战指南本文针对开发者常见的OpenAI API连接错误——HTTPSConnectionPool问题，提供了一套完整的实战排查指南。文章深入分析了错误根源，指出网络出口IP区域限制是关键，并详细讲解了通过基础连通性测试、开关对比法锁定问题，以及针对HTTP代理、SOCKS代理和系统级VPN的不同解决方案，帮助开发者高效修复API连接故障。阅读详情

最近找到githurb上stn的项目，是基于Tensorflow0.7的
本人使用环境 tensorflow1.0 python3.5
以下为报错

File "C:\Users\hasee\STN_tf_test01.py", line 38, in <module>

  h_trans=transformer(x,h_fc1,out_size)

File "C:\Users\hasee\STN_tf_01.py", line 145, in transformer

  output = _transform(theta, U, out_size)

File "C:\Users\hasee\STN_tf_01.py", line 122, in _transform

  grid = _meshgrid(out_height, out_width)

File "C:\Users\hasee\STN_tf_01.py", line 102, in _meshgrid

  grid = tf.concat(0, [x_t_flat, y_t_flat, ones])

File "C:\Users\hasee\AppData\Local\Programs\Python\Python35\Lib\site-packages\tensorflow\python\ops\array_ops.py", line 1047, in concat

  dtype=dtypes.int32).get_shape(

File "C:\Users\hasee\AppData\Local\Programs\Python\Python35\Lib\site-packages\tensorflow\python\framework\ops.py", line 651, in convert_to_tensor

  as_ref=False)

File "C:\Users\hasee\AppData\Local\Programs\Python\Python35\Lib\site-packages\tensorflow\python\framework\ops.py", line 716, in internal_convert_to_tensor

  ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)

File "C:\Users\hasee\AppData\Local\Programs\Python\Python35\Lib\site-packages\tensorflow\python\framework\constant_op.py", line 176, in _constant_tensor_conversion_function

  return constant(v, dtype=dtype, name=name)

File "C:\Users\hasee\AppData\Local\Programs\Python\Python35\Lib\site-packages\tensorflow\python\framework\constant_op.py", line 165, in constant

  tensor_util.make_tensor_proto(value, dtype=dtype, shape=shape, verify_shape=verify_shape))

File "C:\Users\hasee\AppData\Local\Programs\Python\Python35\Lib\site-packages\tensorflow\python\framework\tensor_util.py", line 367, in make_tensor_proto

  _AssertCompatible(values, dtype)

File "C:\Users\hasee\AppData\Local\Programs\Python\Python35\Lib\site-packages\tensorflow\python\framework\tensor_util.py", line 302, in _AssertCompatible

  (dtype.name, repr(mismatch), type(mismatch).__name__))



builtins.TypeError: Expected int32, got list containing Tensors of type '_Message' instead.

报错部分代码如下：



    def _meshgrid(height, width):

        print('begin--meshgrid')

        with tf.variable_scope('_meshgrid'):

            # This should be equivalent to:

            #  x_t, y_t = np.meshgrid(np.linspace(-1, 1, width),

            #                         np.linspace(-1, 1, height))

            #  ones = np.ones(np.prod(x_t.shape))

            #  grid = np.vstack([x_t.flatten(), y_t.flatten(), ones])

            

            x_t = tf.matmul(tf.ones(shape=tf.stack([height, 1])),

                            tf.transpose(tf.expand_dims(tf.linspace(-1.0, 1.0, width), 1), [1, 0]))

            print('meshgrid_x_t_ok')

            y_t = tf.matmul(tf.expand_dims(tf.linspace(-1.0, 1.0, height), 1),

                            tf.ones(shape=tf.stack([1, width])))

            print('meshgrid_y_t_ok')

            x_t_flat = tf.reshape(x_t, (1, -1))

            y_t_flat = tf.reshape(y_t, (1, -1))

            print('meshgrid_flat_t_ok')

            ones = tf.ones_like(x_t_flat)

            print('meshgrid_ones_ok')

            print(x_t_flat)

            print(y_t_flat)

            print(ones)

            grid = tf.concat(0, [x_t_flat, y_t_flat, ones])#报错语句

            print ('over_meshgrid')

            return grid

以及上级代码如下：

im=ndimage.imread('C:\\Users\\hasee\\Desktop\\cat.jpg')

im=im/255.

#im=tf.reshape(im, [1,1200,1600,3])



im=im.reshape(1,1200,1600,3)



im=im.astype('float32')

print('img-over')

out_size=(600,800)

batch=np.append(im,im,axis=0)

batch=np.append(batch,im,axis=0)

num_batch=3



x=tf.placeholder(tf.float32,[None,1200,1600,3])

x=tf.cast(batch,'float32')

print('begin---')

with tf.variable_scope('spatial_transformer_0'):

    n_fc=6

    w_fc1=tf.Variable(tf.Variable(tf.zeros([1200*1600*3,n_fc]),name='W_fc1'))

    initial=np.array([[0.5,0,0],[0,0.5,0]])

    initial=initial.astype('float32')

    initial=initial.flatten()

    

    

    b_fc1=tf.Variable(initial_value=initial,name='b_fc1')

    

    

    h_fc1=tf.matmul(tf.zeros([num_batch,1200*1600*3]),w_fc1)+b_fc1

    

    print(x,h_fc1,out_size)



    h_trans=transformer(x,h_fc1,out_size)

 

    

sess=tf.Session()

sess.run(tf.global_variables_initializer())

y=sess.run(h_trans,feed_dict={x:batch})

plt.imshow(y[0])

plt.show()