如何使用 CNN 代码修复 Python “IndexError: list index out of range”

Question

我正在尝试在 Python 3.7 上使用以下 function 构建一个ResNet34 编码器作为我的 CNN 的一部分。

import tensorflow as tf

from tensorpack import *
from tensorpack.models import BatchNorm, BNReLU, Conv2D, MaxPooling, FixedUnPooling
from tensorpack.tfutils.summary import add_moving_summary, add_param_summary
from .utils import *
import sys

sys.path.append("..") # adds higher directory to python modules path.
try: # HACK: import beyond current level, may need to restructure
    from config import Config
except ImportError:
    assert False, 'Fail to import config.py'
def res_blk(name, l, ch, ksize, count, split=1, strides=1, freeze=False):
    ch_in = l.get_shape().as_list()
    with tf.variable_scope(name):
        for i in range(0, count):
            with tf.variable_scope('block' + str(i)):  
                x = l if i == 0 else BNReLU('preact', l)
                x = Conv2D('conv1', x, ch[0], ksize[0], activation=BNReLU)
                x = Conv2D('conv2', x, ch[1], ksize[1], split=split, 
                                strides=strides if i == 0 else 1, activation=BNReLU)
                x = Conv2D('conv3', x, ch[2], ksize[2], activation=tf.identity)
                if (strides != 1 or ch_in[1] != ch[2]) and i == 0:
                    l = Conv2D('convshortcut', l, ch[2], 1, strides=strides)
                x = tf.stop_gradient(x) if freeze else x
                l = l + x
        l = BNReLU('bnlast',l)  
    return l

def encoder(i, freeze):
    d1 = Conv2D('conv0',  i, 64, 7, padding='valid', strides=1, activation=BNReLU)
    d1 = res_blk('group0', d1, [ 64,  64], [3, 3], 3, strides=1, freeze=freeze)

    d2 = res_blk('group1', d1, [128, 128], [3, 3], 4, strides=2, freeze=freeze)
    d2 = tf.stop_gradient(d2) if freeze else d2

    d3 = res_blk('group2', d2, [256, 256], [3, 3], 6, strides=2, freeze=freeze)
    d3 = tf.stop_gradient(d3) if freeze else d3

    d4 = res_blk('group3', d3, [512, 512], [3, 3], 3, strides=2, freeze=freeze)
    d4 = tf.stop_gradient(d4) if freeze else d4

    d4 = Conv2D('conv_bot',  d4, 1024, 1, padding='same')
    return [d1, d2, d3, d4]

然后我得到错误

 line 67, in encoder
    d1 = res_blk('group0', d1, [ 64,  64], [3, 3], 3, strides=1, freeze=freeze)                       
  File "....", line 34, in res_blk
    x = Conv2D('conv3', x, ch[2], ksize[2], activation=tf.identity)
IndexError: list index out of range

此错误的原因是什么，我该如何解决？ 原始代码是一个运行良好的 Resnet50，即代码是

d1 = res_blk('group0', d1, [ 64, 64, 256], [1, 3, 1], 3, strides=1, freeze=freeze)

Answer 1

我在您包含的链接中发现的关键位是“每个 ResNet 块都是两层深（用于 ResNet 18、34 等小型网络）或 3 层深（ResNet 50、101、152）。”。

您提供的代码适用于 ResNet 50、101 和 152 所需的三层，但据我所知，此示例不适用更简单的两层方法。

可能需要做更多的调整，但您应该能够只删除最后一层，类似于我在下面的内容，否则在 50 层以上的 ResNet 架构中这是必需的。

import tensorflow as tf

from tensorpack import *
from tensorpack.models import BatchNorm, BNReLU, Conv2D, MaxPooling, FixedUnPooling
from tensorpack.tfutils.summary import add_moving_summary, add_param_summary
from .utils import *
import sys

sys.path.append("..") # adds higher directory to python modules path.
try: # HACK: import beyond current level, may need to restructure
    from config import Config
except ImportError:
    assert False, 'Fail to import config.py'
def res_blk(name, l, ch, ksize, count, split=1, strides=1, freeze=False):
    ch_in = l.get_shape().as_list()
    with tf.variable_scope(name):
        for i in range(0, count):
            with tf.variable_scope('block' + str(i)):  
                x = l if i == 0 else BNReLU('preact', l)
                x = Conv2D('conv1', x, ch[0], ksize[0], activation=BNReLU)
                x = Conv2D('conv2', x, ch[1], ksize[1], split=split, 
                                strides=strides if i == 0 else 1, activation=BNReLU)
                x = tf.stop_gradient(x) if freeze else x
                l = l + x
        l = BNReLU('bnlast',l)  
    return l

def encoder(i, freeze):
    d1 = Conv2D('conv0',  i, 64, 7, padding='valid', strides=1, activation=BNReLU)
    d1 = res_blk('group0', d1, [ 64,  64], [3, 3], 3, strides=1, freeze=freeze)

    d2 = res_blk('group1', d1, [128, 128], [3, 3], 4, strides=2, freeze=freeze)
    d2 = tf.stop_gradient(d2) if freeze else d2

    d3 = res_blk('group2', d2, [256, 256], [3, 3], 6, strides=2, freeze=freeze)
    d3 = tf.stop_gradient(d3) if freeze else d3

    d4 = res_blk('group3', d3, [512, 512], [3, 3], 3, strides=2, freeze=freeze)
    d4 = tf.stop_gradient(d4) if freeze else d4

    d4 = Conv2D('conv_bot',  d4, 1024, 1, padding='same')
    return [d1, d2, d3, d4]