ActivityNet数据集下载过程.pdf

ActivityNet数据集下载过程 准备⼯作 ： 安装anaconda pip install youtube-dl sudo apt install ffmpeg pip install pafy 1.⾸先利⽤如下命令，将该 ithub项 ⽬下载下来。或者直接去 ithub上进⾏下载 git clone /activitynet/ActivityNet.git 2.进⼊到Crawler⽂件夹下 4.利⽤下⾯命令创建⼀个存放将要下载的数据集的⽂件夹 mkdir dataset 5.随便建⼀个⼯程，新建⼀个python file⽂档，，将下⾯的代码拷贝进去 6.在ActivityNet官⽹上下载j son⽂件 ： 另保存⽂件 ：activity_net.v 1-3.min.j son 将其复制到Crawler⽂件夹下 7. 将下⾯代码的 directory = dataset/按照你 ⾃⼰的重新要求命名，注意后⾯的 ‘/’不能少。 8. cd 到download.py所在的⽬录，执⾏python 就可以下载啦 import os import j son import uuid import lob import subprocess from j oblib import delayed from j oblib import Parallel # specify download directory directory = dataset/ videoCounter = 0 num_j obs = 24 # open j son file with open(activity_net.v 1-3.min.j son) as data_file: data = j son.load(data_file) # take only video informations from database obj ect videos = data[database] total = len(videos) # iterate throu h dictionary of videos def download_clip(videos, i, key, directory, total): video = videos[key] # find video subset subset = video[subset] # find video label annotations = video[annotations] label = if len(annotations) != 0: label = annotations[0] [label] label = / + label.replace( , _) # create folder named as label if does not exist label_dir = directory + subset + label if not os.path.exists(label_dir): os.makedirs(label_dir) # take url of video url = video[url] tmp_dir = tmp url_base = /watch?v= video_identifier = key tmp_filename = os.path.j oin(tmp_dir, %s.%%(ext)s % uuid.uuid4 ()) command = [youtube-dl, --quiet, --no-warnin s, -f, mp4, -o, %s % tmp_filename, %s % (url_base + video_identifier)] command = .j oin(command) attempts = 0 num_attempts = 5