【可食用系列】lmdb类型数据集生成字典

目标：对包含data.mdb和lock.mdb两个文件的lmdb类型的数据集生成对应的字典，以便后续训练。

import cv2
import lmdb
import numpy as np
import six
from PIL import Image
import os

lmdb_path = '###' #修改成自己的mdb文件所在的文件夹
 
fin = open("./dict.txt","w")
#coding = 'utf8'
coding = ''
env = lmdb.open(lmdb_path)
dict = {}
with env.begin(write=False) as txn:
    #total_number = int(txn.get('total_number'.encode(coding)).decode(coding))
    n_samples = int(txn.get('num-samples'.encode()))
    print(n_samples)
    for index in range(n_samples):
        idx = index + 1
        label_key = 'label-%09d'.encode() % idx
        label = txn.get(label_key).decode()
        img_key = 'image-%09d'.encode() % idx
        imgbuf = txn.get(img_key)
        image_buf = np.frombuffer(imgbuf, dtype=np.uint8)
        for word in label:
            dict[word] = 1

    print(dict.keys())
    print(len(dict))
    for k in dict.keys():
        fin.write(k)
        fin.write('\n')
    fin.close()

然后就会生成一个字典文件，耶！