美文网首页
排行榜视频生产过程大揭秘

排行榜视频生产过程大揭秘

作者: 跑者小越 | 来源:发表于2018-09-24 12:34 被阅读42次

    第一步：获取演奏区当天所有口琴投稿信息

    • 连接上mysql数据库之后使用下列sql语句来获取当天的b站口琴投稿信息。（我们使用的是霜落XXS开发并维护的爬虫数据库）
    SELECT  s.aid,s.title,s.pubdate,up.name,s.mid,s.duration,d.view, s.videos,d.danmaku,d.favorite,d.coin,d.SHARE,d.reply,d.like,d.dislike  FROM 
(SELECT aid, VIEW,danmaku,reply,favorite,coin,SHARE,`like`,dislike FROM video_dynamic_180924 ) AS d, 
(SELECT aid,MID,pubdate,tid,title,duration,videos FROM video_static WHERE ((tid=59) AND (title LIKE '%口琴%') AND(pubdate BETWEEN '2018-08-08' AND '2018-08-15')) )AS s,
(SELECT MID, NAME FROM up_data) AS up 
WHERE d.aid=s.aid AND s.mid = up.mid 
ORDER BY  DATE(pubdate) ASC LIMIT 10000000;

    • 将获取到的数据储存为CSV格式，如下图：


    第二步：计算得分

    • 对于历史投稿，首先计算该月数据与上月数据的差值，然后计算分数。
    • 对于新增投稿，直接计算分数。
    
import time
import sys
import importlib
import math
import pandas as pd 
import numpy as np
import threading
import queue

aid1 = []
title1 = []
name1 = []
view1 = []
score1 = []
mid1 = []
duration1 = []
videos1 = []
danmaku1 = []
favorite1 = []
coin1 = []
share1 = []
reply1 = []
like1 = []
alpha1 = []
pubdate1 = []

df=pd.read_csv('test1.csv', encoding="GBK")
df1 = pd.read_csv('20180923.csv', encoding="GBK")
df2 = pd.read_csv('20180826.csv', encoding="GBK")
df3=pd.merge(df1,df2,on='aid',how='left')
df3.to_csv('20180826-20180923.csv', encoding="GBK")


class myThread (threading.Thread):   
    def __init__(self, thread_id, q):
        threading.Thread.__init__(self)
        self.thread_id = thread_id
        self.q = q

    def run(self):
        while True:
            try:
                t = self.q.get(block=True, timeout=0)
            except queue.Empty:
                break
            try: 
                if  not math.isnan(df3.MID_y[t]) :
                    aid=df3.aid[t]
                    title=df3.title_x[t]
                    pubdate=df3.pubdate_x[t]
                    name=df3.NAME_x[t]
                    mid=df3.MID_x[t]
                    duration=time.strftime("%M:%S", time.localtime(df3.duration_x[t]))
                    videos=df3.videos_x[t]
                    view=df3.VIEW_x[t]-df3.VIEW_y[t]
                    danmaku=df3.danmaku_x[t]-df3.danmaku_y[t]
                    favorite=df3.favorite_x[t]-df3.favorite_y[t]
                    coin=df3.coin_x[t]-df3.coin_y[t]
                    share=df3.SHARE_x[t]-df3.SHARE_y[t]
                    reply=df3.reply_x[t]-df3.reply_y[t]
                    like=df3.like_x[t]-df3.like_y[t]

                    if ((view>0)and(danmaku>=0)and(favorite>=0)and(coin>=0)and(share>=0)and(reply>=0)and(like>=0)):
                        view_fixed=math.sqrt(10000*view)*(4/(videos+3))
                        k=(coin*favorite*400)/(view**2)
                        alpha=k if k<1 else 1
                        favorite_alpha=round(alpha*30, 2)
                        coin_alpha=round(alpha*20, 2)
                        share_alpha=round(alpha*50, 2)
                        reply_alpha=round(alpha*10, 2)
                        score=round(view_fixed+(share*50+favorite*30+coin*20+reply*10)*alpha, 2)
                        aid1.append(aid)
                        title1.append(title)
                        name1.append(name)
                        pubdate1.append(pubdate)
                        view1.append(view)
                        score1.append(score)
                        mid1.append(mid)
                        duration1.append(duration)
                        videos1.append(videos)
                        danmaku1.append(danmaku)
                        favorite1.append(favorite)
                        coin1.append(coin)
                        share1.append(share)
                        reply1.append(reply)
                        like1.append(like)
                        alpha1.append(alpha)   


                            
                else:
                    aid=df3.aid[t]
                    title=df3.title_x[t]
                    pubdate=df3.pubdate_x[t]
                    name=df3.NAME_x[t]
                    mid=df3.MID_x[t]
                    duration=time.strftime("%M:%S", time.localtime(df3.duration_x[t]))
                    videos=df3.videos_x[t]
                    view=df3.VIEW_x[t]
                    danmaku=df3.danmaku_x[t]
                    favorite=df3.favorite_x[t]
                    coin=df3.coin_x[t]
                    share=df3.SHARE_x[t]
                    reply=df3.reply_x[t]
                    like=df3.like_x[t]


                    view_fixed=math.sqrt(10000*view)*(4/(videos+3))
                    k=(coin*favorite*400)/(view**2)
                    alpha=k if k<1 else 1
                    favorite_alpha=round(alpha*30, 2)
                    coin_alpha=round(alpha*20, 2)
                    share_alpha=round(alpha*50, 2)
                    reply_alpha=round(alpha*10, 2)
                    score=round(view_fixed+(share*50+favorite*30+coin*20+reply*10)*alpha, 2)
                    aid1.append(aid)
                    title1.append(title)
                    name1.append(name)
                    pubdate1.append(pubdate)
                    view1.append(view)
                    score1.append(score)
                    mid1.append(mid)
                    duration1.append(duration)
                    videos1.append(videos)
                    danmaku1.append(danmaku)
                    favorite1.append(favorite)
                    coin1.append(coin)
                    share1.append(share)
                    reply1.append(reply)
                    like1.append(like)
                    alpha1.append(alpha)
                        

                print(t)

                
            except:
                pass     

q = queue.Queue()



#把JOBS排入队列
for i in range(0,df3.shape[0]-1):
    q.put(i)

n = 16
thread_list = [myThread(i, q) for i in range(n)]


#fork NUM个线程等待队列
for i in range(n):
    thread_list[i].start()

#等待所有JOBS完成
for i in range(n):
    thread_list[i].join()
    
    


columns = {'aid': aid1, 'title': title1, 'pubdate':pubdate1,'name': name1,'mid':mid1,'duration':duration1,'view': view1,'danmaku':danmaku1,'favorite':favorite1,'coin':coin1,'share':share1,'reply':reply1,'like':like1,'alpha':alpha1,'score':score1 }
df = pd.DataFrame(columns)
df = df[['aid','title','pubdate','name','mid','duration','view','danmaku','favorite','coin','share','reply','like','alpha','score']]
df.to_csv('【20180826-20180923】.csv', encoding="GBK")

    • 处理后的数据如下：


    第三步：制作模板图片

    首先准备好空白模板


    • 按照排名往里面填充对应数据
    from PIL import Image,ImageFont,ImageDraw #引入相关库
import pandas as pd

df=pd.read_csv('【20180826-20180923】.csv', encoding="GBK")
i=0
##前十

while i<10:
    t=i+1
    k1="NEW !!" if df.pubdate[i] > '2018/8/26 00:00' else "Old !!"
    k3=df.duration[i]
    k4=df.score[i]
    k5=df.aid[i]
    k6=df.pubdate[i]
    k7=df.title[i]
    k8=df.name[i]
    k9=df.reply[i]
    k10=df.view[i]
    k11=df.favorite[i]
    k12=df.share[i]
    k13=df.coin[i]

    font1=ImageFont.truetype('1.TTF',150) #名次
    font2=ImageFont.truetype('BRDWAYN.TTF',60) #NEW&Old
    font3=ImageFont.truetype('1.TTF',60) #time
    font4=ImageFont.truetype('1.TTF',50) #right-white
    font5=ImageFont.truetype('1.TTF',40)



    def mytext(self, pos, text, font, fill, border='black', bp=1):
            x, y = pos
            shadowcolor = border
            self.text((x-bp, y), text, font=font, fill=shadowcolor)
            self.text((x+bp, y), text, font=font, fill=shadowcolor)
            self.text((x, y-bp), text, font=font, fill=shadowcolor)
            self.text((x, y+bp), text, font=font, fill=shadowcolor)
            # thicker border
            self.text((x-bp, y-bp), text, font=font, fill=shadowcolor)
            self.text((x+bp, y-bp), text, font=font, fill=shadowcolor)
            self.text((x-bp, y+bp), text, font=font, fill=shadowcolor)
            self.text((x+bp, y+bp), text, font=font, fill=shadowcolor)
            # now draw the text over it
            self.text((x, y), text, font=font, fill=fill)


    im=Image.open('10.png')
    draw=ImageDraw.Draw(im)


    mytext(draw, (1580,20), '{}'.format(t), font=font1, fill='red', border='white', bp=4)   # order
    mytext(draw, (1610,185), '{}'.format(k1), font=font2, fill=(255,240,0), border='white', bp=0)   # NEW
    mytext(draw, (1610,300), '{}'.format(k3), font=font3, fill='white', border='white', bp=0)   # time
    mytext(draw, (1170,873), '{}'.format(k4), font=font4, fill='red', border='white', bp=2) # score
    mytext(draw, (230,880), '{}'.format(k5), font=font5, fill='white', border='white', bp=0) # aid
    mytext(draw, (620,880), '{}'.format(k6), font=font5, fill='white', border='white', bp=0) # date
    mytext(draw, (100,965), '{}'.format(k7), font=font5, fill='black', border='black', bp=0) # title
    mytext(draw, (620,1035), '{}'.format(k8), font=font5, fill='white', border='white', bp=0) # up
    mytext(draw, (1750,484), '{}'.format(k9), font=font4, fill='white', border='white', bp=0) # reply
    mytext(draw, (1750,638), '{}'.format(k12), font=font4, fill='white', border='white', bp=0) # share
    mytext(draw, (1710,405), '{}'.format(k10), font=font4, fill='black', border='black', bp=0) # view
    mytext(draw, (1710,560), '{}'.format(k11), font=font4, fill='black', border='black', bp=0) # favorite
    mytext(draw, (1710,715), '{}'.format(k13), font=font4, fill='black', border='black', bp=0) # coin

    im.save('target{}.png'.format(i) )
    i+=1

    • 得到最终图片


    • 副榜同理
    ##副榜31-110

i=0

while i<1:


    font1=ImageFont.truetype('1.TTF',150) #名次
    font2=ImageFont.truetype('BRDWAYN.TTF',30) #NEW&Old
    font3=ImageFont.truetype('1.TTF',60) #time
    font4=ImageFont.truetype('1.TTF',50) #right-white
    font5=ImageFont.truetype('1.TTF',40)
    font6=ImageFont.truetype('1.TTF',30)
    font7=ImageFont.truetype('1.TTF',20)



    def mytext(self, pos, text, font, fill, border='black', bp=1):  #字体描边函数
            x, y = pos
            shadowcolor = border
            self.text((x-bp, y), text, font=font, fill=shadowcolor)
            self.text((x+bp, y), text, font=font, fill=shadowcolor)
            self.text((x, y-bp), text, font=font, fill=shadowcolor)
            self.text((x, y+bp), text, font=font, fill=shadowcolor)
            # thicker border
            self.text((x-bp, y-bp), text, font=font, fill=shadowcolor)
            self.text((x+bp, y-bp), text, font=font, fill=shadowcolor)
            self.text((x-bp, y+bp), text, font=font, fill=shadowcolor)
            self.text((x+bp, y+bp), text, font=font, fill=shadowcolor)
            # now draw the text over it
            self.text((x, y), text, font=font, fill=fill)


    im=Image.open('110.png')
    draw=ImageDraw.Draw(im)
    j=0
    
    while j<8:
        t=30+(i)*8+j+1
        k1="NEW !!" if df.pubdate[i] > '2018/8/26 00:00' else ""
        k3=df.duration[30+(i)*8+j+1]
        k4=df.score[30+(i)*8+j+1]
        k5='av{}'.format(df.aid[30+(i)*8+j+1])
        k6=df.pubdate[30+(i)*8+j+1]
        k7=df.title[30+(i)*8+j+1]
        k8=df.name[30+(i)*8+j+1]
        k9=df.reply[30+(i)*8+j+1]
        k10='播放增量：{}'.format(df.view[30+(i)*8+j+1])
        k11='收藏增量：{}'.format(df.favorite[30+(i)*8+j+1])
        k12=df.share[30+(i)*8+j+1]
        k13='硬币增量：{}'.format(df.coin[30+(i)*8+j+1])

        mytext(draw, (1770,80+120*j), '{}'.format(t), font=font3, fill='red', border='white', bp=4)   # 名次
        mytext(draw, (1835,90+120*j), '{}'.format(k1), font=font2, fill='green', border='white', bp=0)   # NEW
        # mytext(draw, (1650,185), '{}'.format(k2), font=font2, fill=(255,240,0), border='white', bp=0) 
#        mytext(draw, (630,130+120*j), '{}'.format(k3), font=font7, fill='black', border='black', bp=0)   # time
        mytext(draw, (1600,90+120*j), '{}'.format(k4), font=font5, fill='black', border='black', bp=0) #score
        mytext(draw, (430,130+120*j), '{}'.format(k5), font=font7, fill='black', border='black', bp=0) #aid
        mytext(draw, (750,130+120*j), '{}'.format(k6), font=font7, fill='black', border='black', bp=0) #date
        mytext(draw, (420,80+120*j), '{}'.format(k7), font=font6, fill='black', border='black', bp=0) #title
        mytext(draw, (540,130+120*j), '{}'.format(k8), font=font7, fill='black', border='black', bp=0) #up
#         mytext(draw, (1750,484), '{}'.format(k9), font=font4, fill='white', border='white', bp=0) #评论
#         mytext(draw, (1750,638), '{}'.format(k12), font=font4, fill='white', border='white', bp=0) #分享
        mytext(draw, (1000,130+120*j), '{}'.format(k10), font=font7, fill='green', border='green', bp=0) #播放
        mytext(draw, (1200,130+120*j), '{}'.format(k11), font=font7, fill='green', border='green', bp=0) #收藏
        mytext(draw, (1400,130+120*j), '{}'.format(k13), font=font7, fill='green', border='green', bp=0) #硬币
        j+=1
    im.save('target{}.png'.format(30+i) )
    i+=1

    副榜效果图


    第四步：选取视频片段

    • 下载视频


      image.png

    • 选取片段（一般为高潮部分）

    第五步：剪辑视频

    • 用PR剪辑视频


    • 渲染生成最终作品

    相关文章

      网友评论

          本文标题：排行榜视频生产过程大揭秘

          本文链接：https://www.haomeiwen.com/subject/ckmcoftx.html

          延伸阅读

          深度阅读

          • 您也可以注册成为美文阅读网的作者，发表您的原创作品、分享您的心情！

          栏目导航

          热点阅读

          关于我们|服务条款|联系我们|排行榜视频生产过程大揭秘|投稿指南|网站地图|RSS订阅|排版工具|手机版

          提供经典美文摘抄,优美散文欣赏,现代诗歌精选,短篇小说,心情随笔,表白情书范文,故事会在线阅读欣赏

          Copyright © 2014-2023 Haomeiwen.com All Rights Reserved. 好美文阅读网 版权所有

          备案信息：桂公网安备 45052102000051号 · 桂ICP备13007215号-3

          本站所收录作品、热点评论等信息部分来源互联网，目的只是为了系统归纳学习和传递资讯

          所有作品版权归原创作者所有，与本站立场无关，如不慎侵犯了你的权益，请联系我们告知，我们将做删除处理！