使用Python对Tensorflow的计算结果进行简单分析

项目位置：Python
代码位置：DataAnalysis.py
项目Python的data目录带有需要分析的数值

file = './data/neutral.txt'

#这些是已知的分类
names = ['drawings', 'hentai', 'neutral', 'porn', 'sexy']
itemList = []

with open(file) as f:
    item = {}

    for line in f.readlines():
        line = line.strip('\n')
        #每个以'.jpg'结尾的行，作为一个item的开始
        if line.endswith('.jpg'):
            if item != {}:
                t = item.copy()
                itemList.append(t)
                item.clear()
            item['pic'] = line
        else:
            splits = line.split(' ')
            if len(splits) > 1:
                key = splits[0].strip()
                value = splits[3].strip(')')
                item[key] = value
    #将最后一item放入队列
    if item != {}:
        itemList.append(item)

maxList = []
itemNumber = len(itemList)
for i in itemList:
    pic = i.get('pic')
    maxKey = max(i, key=i.get)
    # if(maxKey == 'porn'):
    #     print(pic + " : " + str(maxKey))
    maxList.append(maxKey)

for n in names:
    print('%8s : %d' % (n, maxList.count(n)))
    #print(n + ' : ' + str(maxList.count(n)))

需要分析的内容是这样的：

./test/hentai/4b8023e0-c673-4a6f-8589-603ee9d93855.jpg
hentai (score = 0.82976)
drawings (score = 0.07327)
porn (score = 0.05042)
sexy (score = 0.04333)
neutral (score = 0.00321)

./test/hentai/2b383b71-b72b-4d71-8fce-a64cbc604eea.jpg
hentai (score = 0.47422)
drawings (score = 0.34337)
neutral (score = 0.12827)
porn (score = 0.03460)
sexy (score = 0.01954)

./test/hentai/331da44a-5cbf-40dc-b91e-8ba91e652875.jpg
hentai (score = 0.70981)
drawings (score = 0.22027)
porn (score = 0.03091)
sexy (score = 0.02160)
neutral (score = 0.01741)

打印结果：

drawings : 77
  hentai : 10
 neutral : 1783
    porn : 77
    sexy : 53

这样就对分类有了一个大概的了解。