&figure&&img src=&https://pic3.zhimg.com/v2-4ca03a3aeeff2b2f30f65be_b.jpg& data-rawwidth=&300& data-rawheight=&198& class=&content_image& width=&300&&&/figure&&br&&br&&br&&br&&figure&&img src=&https://pic2.zhimg.com/v2-aa26f88c8adff4f_b.jpg& data-rawwidth=&444& data-rawheight=&210& class=&origin_image zh-lightbox-thumb& width=&444& data-original=&https://pic2.zhimg.com/v2-aa26f88c8adff4f_r.jpg&&&/figure&&br&&br&&br&&br&&br&&figure&&img src=&https://pic3.zhimg.com/v2-70e19441b4ebd559ea2152_b.jpg& data-rawwidth=&233& data-rawheight=&187& class=&content_image& width=&233&&&/figure&&br&&br&&br&&br&&br&&figure&&img src=&https://pic1.zhimg.com/v2-3db403e88_b.jpg& data-rawwidth=&300& data-rawheight=&268& class=&content_image& width=&300&&&/figure&&br&&br&&br&&br&&br&&br&&figure&&img src=&https://pic3.zhimg.com/v2-657a871cadc_b.jpg& data-rawwidth=&430& data-rawheight=&215& class=&origin_image zh-lightbox-thumb& width=&430& data-original=&https://pic3.zhimg.com/v2-657a871cadc_r.jpg&&&/figure&&br&&br&&figure&&img src=&https://pic4.zhimg.com/v2-fc569bdff8f_b.jpg& data-rawwidth=&250& data-rawheight=&400& class=&content_image& width=&250&&&/figure&&br&&br&&figure&&img src=&https://pic1.zhimg.com/v2-ae3eafa2e08_b.jpg& data-rawwidth=&227& data-rawheight=&197& class=&content_image& width=&227&&&/figure&&br&&br&&figure&&img src=&https://pic4.zhimg.com/v2-19de13da24a105f67a0933_b.jpg& data-rawwidth=&280& data-rawheight=&370& class=&content_image& width=&280&&&/figure&&br&&br&&figure&&img src=&https://pic1.zhimg.com/v2-b898751accaf_b.jpg& data-rawwidth=&245& data-rawheight=&286& class=&content_image& width=&245&&&/figure&&br&&br&&figure&&img src=&https://pic1.zhimg.com/v2-21daab0eb6518aef195c3f5e_b.jpg& data-rawwidth=&170& data-rawheight=&199& class=&content_image& width=&170&&&/figure&&br&&br&&figure&&img src=&https://pic1.zhimg.com/v2-e4b1ed6dc_b.jpg& data-rawwidth=&300& data-rawheight=&169& class=&content_image& width=&300&&&/figure&&br&&br&&figure&&img src=&https://pic2.zhimg.com/v2-fa5da3efababb86316b25_b.jpg& data-rawwidth=&240& data-rawheight=&320& class=&content_image& width=&240&&&/figure&&br&&br&&figure&&img src=&https://pic4.zhimg.com/v2-748ff204d2ae4a7f859b3_b.jpg& data-rawwidth=&350& data-rawheight=&197& class=&content_image& width=&350&&&/figure&&br&&br&&figure&&img src=&https://pic1.zhimg.com/v2-d7dc05fdf2e_b.jpg& data-rawwidth=&301& data-rawheight=&323& class=&content_image& width=&301&&&/figure&&br&&br&&figure&&img src=&https://pic3.zhimg.com/v2-563cf5dbbc79eab74acf6bf823031efa_b.jpg& data-rawwidth=&240& data-rawheight=&320& class=&content_image& width=&240&&&/figure&&br&&br&&figure&&img src=&https://pic2.zhimg.com/v2-f99b102fa97baacae909_b.jpg& data-rawwidth=&152& data-rawheight=&168& class=&content_image& width=&152&&&/figure&&br&&br&&figure&&img src=&https://pic4.zhimg.com/v2-490b69bf854c7ebf_b.jpg& data-rawwidth=&350& data-rawheight=&311& class=&content_image& width=&350&&&/figure&&br&&br&&figure&&img src=&https://pic4.zhimg.com/v2-e6ba1db5f_b.jpg& data-rawwidth=&200& data-rawheight=&152& class=&content_image& width=&200&&&/figure&&br&&br&&figure&&img src=&https://pic2.zhimg.com/v2-e44c7c1ee027d09d8c0d_b.jpg& data-rawwidth=&320& data-rawheight=&238& class=&content_image& width=&320&&&/figure&&br&&br&&figure&&img src=&https://pic3.zhimg.com/v2-3c39a7ffe83f95f5d6ee5a_b.jpg& data-rawwidth=&350& data-rawheight=&227& class=&content_image& width=&350&&&/figure&
&p&破8千赞，回答上了知乎日报啦～已删去鸡肋推荐并增加了更为地道的美食，剩下的都是集锦。&/p&&p&大学四年在苏州，谈一谈眼里的苏州好去处和好味道。&/p&&p&&br&&/p&&p&&b&旅行必体验之景点篇：&/b&&/p&&p&&br&&/p&&p&&b&1 苏州大学看老式红楼和钟楼&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/c7d6eaefa64dc4e796a3b8_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&2144& data-rawheight=&1424& class=&origin_image zh-lightbox-thumb& width=&2144& data-original=&https://pic1.zhimg.com/c7d6eaefa64dc4e796a3b8_r.jpg&&&/figure&&p&博物馆和大学是我认为一个城市的精髓所在之一，苏州大学本部就是这样一个城市之眼所在。非常有历史感的一个校区，集合了苏州的园林和水乡元素，自然与人文相互交融。苏大不像厦大一样游人如织，喜欢这里更多是因为这里有着难能可贵的清幽。民国风味十足，前身是光绪年间的东吴大学。&/p&&p&红楼和钟楼，也就是数科院和校长室是校园内必看的两处。天气好的午后，学生们多会集中在大草坪和情人坡，聆听他们弹吉他、观望他们踢足球、欣赏他们画速写......无论随便一处都一定会让你留恋其中。&/p&&p&Tips：苏大本部在平江区十梓街1号，不要去成了独墅湖校区的。现在苏大管理很严格，一定要带上各种证件，然后再打扮地尽量像学生一些。小道消息可以参考&a href=&https://www.zhihu.com/question/& class=&internal&&苏州大学本部（天赐庄）校区为什么不对外开放？ - 大学&/a&这个回答。&/p&&p&&br&&/p&&p&&b&2 苏州博物馆赏贝聿铭的雅致设计&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/170d6cbd3c0ba2e15a28_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&800& data-rawheight=&528& class=&origin_image zh-lightbox-thumb& width=&800& data-original=&https://pic1.zhimg.com/170d6cbd3c0ba2e15a28_r.jpg&&&/figure&&p&知道贝聿铭的一定会来拜访这里，建筑大师设计的作品，中国色彩和苏州风情处处可见。既保留了苏州古建筑中白墙灰瓦的风格，又融入了现代的几何线条。中国的建筑多见的是繁复的美，但是苏博更多的是一种诗情画意的柔美。馆中还有庭院、水池、石桥，既传统又现代。&/p&&p&当然除了建筑之外，管内的藏品也非常值得细细品味。镇馆之宝“越窑秘色瓷莲花碗”、“真珠舍利宝幢”可以特别留意。&/p&&p&Tips:苏博的官网可以参观时间，免去长时间的排队等候。每张有效身份证件每天限定预约一次，最多实名预约3人。预约成功后再预约时间来参观，可走绿色快速通道。&/p&&p&&br&&/p&&p&&br&&/p&&p&&b&3 平江路看小桥流水 逛文艺古宅&/b&&/p&&figure&&img src=&https://pic3.zhimg.com/41b5e3f68ae4147cdd59d1661a1eefb2_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&684& data-rawheight=&419& class=&origin_image zh-lightbox-thumb& width=&684& data-original=&https://pic3.zhimg.com/41b5e3f68ae4147cdd59d1661a1eefb2_r.jpg&&&/figure&&p&越来越多的旅游城市都有这样一条街巷，沉淀下历史的味道又融合进城市的新气息，形成创意文化街区。上海的田子坊、北京的南锣鼓巷、扬州的东关街......苏州的平江路更是美食聚集地、创意文化街和古镇风光三者相结合。&/p&&p&岸边绿树环绕，偶尔船工摇橹高歌；小吃、菜馆、文创店，商铺林立；水陆并行、河街相邻。&/p&&p&Tips：和山塘街相比，同样是“古镇风光”，但更推荐平江路。平江路散客多，山塘街团客多。如果两个都想一探究竟的，建议早上来平江路，人少清静；晚上去山塘街，看夜色迷离。&/p&&p&&b&4 苏州火车站观“最美火车站”&/b&&/p&&figure&&img src=&https://pic3.zhimg.com/8ffdc559b1cc74249bcc696e_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&399& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic3.zhimg.com/8ffdc559b1cc74249bcc696e_r.jpg&&&/figure&&p&&br&&/p&&p&苏州火车站是建筑设计大师崔恺担纲主创设计的，个人认为国内火车站设计最具有地方特色的一个火车站，已入选苏州十大建筑。苏州火车站的设计，给人的主要印象在于整体上是各种几何线条组成的菱形、方形、圆形的粉墙黛瓦世界。火车站南就是护城河，隔河就是古城区，北寺塔、平门、桃花坞近在咫尺，在保证大体量的现代化交通设施能够自然承接、延续古城的肌理上，苏州火车站的设计还是可圈可点的。&/p&&p&Tips：是苏州火车站，不是苏州园区站。&/p&&p&&br&&/p&&p&&b&5 天平山赏红叶 道天凉好个秋&/b&&/p&&figure&&img src=&https://pic4.zhimg.com/cdea509dcaba64eed457496b_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&398& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic4.zhimg.com/cdea509dcaba64eed457496b_r.jpg&&&/figure&&p&&br&&/p&&p&苏州的天平山为中国四大赏红叶胜地之一，以怪石、清泉、红枫“三绝”著称，尤以红枫为最，每年初冬，山中红叶遍野，景色美不胜收。山东南麓之古枫林，为明代万历年间宋代名臣范仲淹十七世孙范允临从福建带回，植于这主莹之地。现存158棵，深秋时节，满山红叶若红霞缭绕，有“天平山红枫甲天下”之誉。也因为这个美名，这个季节很多摄影爱好者慕名而来拍摄枫叶，自然就热闹了。&/p&&p&Tips：天平山离市区较远，注意交通时间的安排；看红枫叶的最佳时间是11月底到十二月初，此期间尽量避免周末前往，人满为患。&/p&&p&&br&&/p&&p&&b&6 太湖观湖光山色 品“太湖三白”&/b&&/p&&figure&&img src=&https://pic2.zhimg.com/ed275f2d99_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&397& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic2.zhimg.com/ed275f2d99_r.jpg&&&/figure&&p&大学期间有次和朋友自驾去太湖，开车吹着风，渔人撒出夕阳下的网，然后在一个绿意盎然的饭店吃白虾，着实难忘。&/p&&p&太湖地跨苏州、无锡、常州、湖州四个大市，包含众多风景区，主要的景色集中在苏州的西山景区和东山景区、无锡的蠡湖景区。全国五大淡水湖之一的太湖，距离苏州市区比较远，大概50多公里，太湖一天没法玩得完，至少得两天时间。当年西施与范蠡泛舟于太湖之上，太湖之美可想而知不是浪得虚名。来这不能错过“太湖三白”—白鱼、银鱼和白虾，当然入秋了还有美味的大闸蟹。&/p&&p&Tips：到达太湖的公共交通最好是从苏州火车站坐69路再转697路公交去到太湖西山景区，车程顺畅的话大概2个半小时就可以到了。&/p&&p&如果觉得太湖太远交通不便，也可以去吴中区看一看石湖。&a class=&member_mention& href=&//www.zhihu.com/people/fc9f8cb607df41e4fc33ad12& data-hash=&fc9f8cb607df41e4fc33ad12& data-hovercard=&p$b$fc9f8cb607df41e4fc33ad12&&@peppps&/a& ：强烈推荐到古城西南角的上方山脚下看看苏州的西湖。有西湖2/3面积的&b&石湖&/b&，范成大，杨万里和姜白石在这里聚会隐居，乾隆数次到这里驻跸行宫。石湖湖山相映如画，园林，岛屿，长堤，宝塔，样样不缺，古意盎然，十里青山，一湖碧水，湖滨繁华都市，宛如西湖。&/p&&p&&br&&/p&&p&&b&7 诚品书店闻书香 逛艺文空间&/b&&/p&&figure&&img src=&https://pic4.zhimg.com/a507afb6bfb3ed6bea437_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&396& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic4.zhimg.com/a507afb6bfb3ed6bea437_r.jpg&&&/figure&&p&&br&&/p&&p&苏州的这家诚品是大陆首家诚品书店，虽然诚品以 24 小时书店出名，事实上书店背后是一整套集合了生活方式、艺文表演的综合商业模式，苏州诚品也不例外。设计师姚仁喜将其定位于“一座人文阅读、创意探索的美学生活博物馆”，总面积达到 56000 平方米的空间划分为地上三层及地下一层，安放下五类书区、四处特色空间、三家店中店以及 B1 层的诚品生活采集。&/p&&p&在网络上看到一句话“今天的苏州，不是挤在诚品就是在去诚品的路上”，诚品还是不要周末去的好，不止是书店的书店，找个工作日的时间去一探究竟。&/p&&p&Tips：可以预留半天至一天的时间在这里，书店很大，里面的配套设施也很齐全。&/p&&p&&br&&/p&&p&&b&8 赏中国四大园林之一的留园 看亭台水榭&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/85ad778b90c0ee9c98f8_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&400& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic1.zhimg.com/85ad778b90c0ee9c98f8_r.jpg&&&/figure&&p&大一刚入校老师就建议我们办园林卡，苏州园林太多，可以花时间一个个慢慢看。拙政园、狮子林、网师园......这么多个园林，为什么我推荐去留园？大家都去拙政园，却只有很少的人去留园。拙政园胜在大，一层又一层的园子，但感觉没有居家的气息。留园就不同了，虽没拙政园大，但也是移步换景，而且处处可以联想到古人的生活场景。关键是留园人少很多（感谢 &a class=&member_mention& href=&//www.zhihu.com/people/015d7ee53d7b& data-hash=&015d7ee53d7b& data-hovercard=&p$b$015d7ee53d7b&&@小旅&/a& 提醒：这里的人少是相对于其他园林而言，任何景点都有平高峰期的），可以细细看景。中国四大园林之一的留园，到处都是亭台水榭，如诗如画。&/p&&p&Tips：留园不远处就是虎丘，可以一并游览。（如果对历史不是太感兴趣的可以略过虎丘）&/p&&p&&a class=&member_mention& href=&//www.zhihu.com/people/fc9f8cb607df41e4fc33ad12& data-hash=&fc9f8cb607df41e4fc33ad12& data-hovercard=&p$b$fc9f8cb607df41e4fc33ad12&&@peppps&/a& ：苏州古典园林在古城足足有三十多座，集成了宋元明以及民国建筑园林艺术风格，各有特色，强烈推荐去看看&b&耦园，沧浪亭，可园，艺圃，网师园，五峰园，环秀山庄，怡园，北半圆，曲园，桃园，天香小筑&/b&，朴园这些个星星点点散落在古城里的精雅园林，它们并不比拙政园，狮子林，虎丘和留园来得逊色。&/p&&p&&br&&/p&&p&&b&9 夜游李公堤 踏苏州最美夜景地&/b&&/p&&figure&&img src=&https://pic3.zhimg.com/92f9e900e28058fcfa9c6810feaf9532_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&514& data-rawheight=&308& class=&origin_image zh-lightbox-thumb& width=&514& data-original=&https://pic3.zhimg.com/92f9e900e28058fcfa9c6810feaf9532_r.jpg&&&/figure&&p&苏州看夜景推荐地，位于工业园区，占地面积极大，环境优美。李公堤是苏州金鸡湖中唯一的湖中长堤，建筑采用中西结合的方式建造，因此，身处这里，既有欧洲大道的静谧和树影婆娑，又有江南建筑的灵气秀美。李公堤有很多美术馆、艺术展览馆，文艺气息在这里的空气中流动，除了夜景白天来这里也是不错的选择。汇集了很多美食，你可以不出李公堤就享受到泰餐、印度料理、苏帮菜、意大利菜、日韩料理甚至素斋。秋天的时候，当碧水印着秋阳，凉风吹落树叶时，和爱人孩子漫步在这里，似乎周围的一切都安静舒适下来了。&/p&&p&Tips：这里有免费的小火车坐，小朋友一定会很喜欢，小火车基本是绕一圈的路线，车次不少，有去也有回，单程十几二十分钟。&/p&&p&&a class=&member_mention& href=&//www.zhihu.com/people/bae69e1cf09a0dc23b7df& data-hash=&bae69e1cf09a0dc23b7df& data-hovercard=&p$b$bae69e1cf09a0dc23b7df&&@方cham bo&/a& ：我觉得要安利一个现代的风景，&b&苏州文化博览中心&/b&出来 秋天阳光正好的时候往&b&玲珑湾&/b&，秋裤门的方向看 倚着栏杆 吹着微风 湖上涟漪就像绸缎一样轻柔 身旁是慵懒的野猫趴在桌上 心一下子就能静下来。（视角最好的地方已经造了星巴克，感觉不太完美了）&/p&&p&&br&&/p&&p&&b&10 沙家浜游览湿地风光 吃阳澄湖外的正宗大闸蟹&/b&&/p&&figure&&img src=&https://pic2.zhimg.com/v2-c4e0d66bf03d0_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&375& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic2.zhimg.com/v2-c4e0d66bf03d0_r.jpg&&&/figure&&p&&br&&/p&&p&相信你对苏州阳澄湖的大闸蟹早有耳闻，沙家浜就位于阳澄湖的北侧，有点杭州西溪湿地的意境。京剧样板戏《沙家浜》的发源地，电视剧《沙家浜》也在此拍摄。坐乌篷船深入芦苇荡，游览湿地风光。春季绿意盎然，一片静谧；夏季荷花盛开，一片恬雅；秋冬芦苇枯黄，一片萧索。无论是哪个季节来，都各有一番滋味。&/p&&p&还有很多表演可看：红石民俗文化村的古戏台可看京剧《沙家浜》著名唱段、横泾老街影视基地的戏台、同样有京剧《沙家浜》可看、横泾老街上还有挑花担表演，周末则有婚俗表演可看。以上表演都是免费的。实景剧《芦荡烽火》更是颇受好评，效果逼真，在横泾老街北侧的芦荡剧场看。另外一场诙谐的实景剧《让子弹飞》在刁宅大院东边的横泾剧场看。&/p&&p&Tips：美景肯定还要配上美食，来这里吃大闸蟹可以说是阳澄湖外“另辟小径”，大闸蟹也是绝对正宗味道极其鲜美的。&/p&&p&&br&&/p&&p&&b&11 古城河游船 叹运河一代风华&/b&&/p&&figure&&img src=&https://pic4.zhimg.com/87ab0d95010ddd20ed1fe170a7e0e577_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&904& data-rawheight=&562& class=&origin_image zh-lightbox-thumb& width=&904& data-original=&https://pic4.zhimg.com/87ab0d95010ddd20ed1fe170a7e0e577_r.jpg&&&/figure&&p&&br&&/p&&p&&a class=&member_mention& href=&//www.zhihu.com/people/fc9f8cb607df41e4fc33ad12& data-hash=&fc9f8cb607df41e4fc33ad12& data-hovercard=&p$b$fc9f8cb607df41e4fc33ad12&&@peppps&/a& ：此外，强烈推荐到石路神仙庙万人码头乘坐大团购三十来元的&b&环古城运河游览船&/b&，一路环游古城，清澈宽阔的春秋伍子胥挖掘的环城河，完整的阊门，金门，胥门，盘门，蛇门，相门，娄门，平门以及葑门和齐门遗址，大片绿化犹如绿带环绕着14.6平方公里的国内最大古城，古韵绿意，美不胜收。&/p&&p&&br&&/p&&p&&b&行程线路推荐篇：&/b&&/p&&p&私信里偶尔会收到一些小伙伴要推荐线路的，这边整一个我认为合理的推荐，5-6天基本上可以有个苏州深度游了。&/p&&p&Day1 苏州博物馆—拙政园—平江路—观前街&/p&&p&Day2 诚品书店—金鸡湖&/p&&p&Day3 苏州大学—留园—山塘街—古运河游船&/p&&p&Day4 太湖/天平山（这两个地方都比较远，自驾的朋友可以考虑，天平山作为游客11月底12月初才建议去）&/p&&p&Day5 同里/周庄/甪直/沙家浜/阳澄湖（古镇或者吃大闸蟹，都是很有苏意的一个体验，时间不够选择一个就好。古镇喜欢热闹的最推荐周庄）&/p&&p&&br&&/p&&p&&b&旅行必体验之美食篇：&/b&&/p&&p&&br&&/p&&p&&b&&u&（一）最有苏味最老字号的苏州早点：&/u&&/b&&/p&&p&&br&&/p&&p&&b&01 双塔点心铺&/b&&/p&&figure&&img src=&https://pic2.zhimg.com/07f005bd4ccda85ea73399_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&399& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic2.zhimg.com/07f005bd4ccda85ea73399_r.jpg&&&/figure&&p&他家总共就卖韭菜蛋饼、葱油饼、豆浆、豆腐花等几种点心，蛋饼和葱油饼的油酥都很足，吃起来香酥可口也不会觉得油腻，豆腐花和豆浆味道很纯正。很多爷爷奶奶都会坐在小小的店铺里一手一个饼，一手喝着豆浆或是豆腐花，还能遇到很多从新区、园区赶来的死忠粉。&/p&&p&推荐：豆腐花、葱油饼&/p&&p&地址：凤凰街双塔菜市场大门处&/p&&p&&br&&/p&&p&&b&02 老西白点心店&/b&&/p&&figure&&img src=&https://pic3.zhimg.com/15a1a9a6_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&420& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic3.zhimg.com/15a1a9a6_r.jpg&&&/figure&&p&他家的生煎和汤圆最好吃，也有人觉得他家的大馄饨最好吃，还有人强推他家的小笼，美食本来就是因人而异，不过可以说明他家这些东西都不错。开了几十年的老字号，生意一直很好。&/p&&p&推荐：生煎、汤圆 &/p&&p&地址：白塔西路104号（近接驾桥）&/p&&p&&br&&/p&&p&&b&03 新年点心店&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/1fbab334bff9f6f0a291b4_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&434& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic1.zhimg.com/1fbab334bff9f6f0a291b4_r.jpg&&&/figure&&p&吃过他家汤团的人，都说吃到了过去苏州汤团的感觉。皮子黏黏QQ的，一口咬下去芝麻馅儿流出来，就像拍广告一样。鲜肉馅儿的汤团也是汤汁满满，汤团都是现包的，按照苏州人的吃法，早饭一碗小馄饨两个汤团，幸福一整天。&/p&&p&推荐：鲜肉汤团、小馄饨&/p&&p&地址：山塘街351号（近星桥堍)&/p&&p&&br&&/p&&p&&b&04 荣阳楼 &/b&&/p&&figure&&img src=&https://pic2.zhimg.com/ff248e7a597f5f7c9db9_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&409& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic2.zhimg.com/ff248e7a597f5f7c9db9_r.jpg&&&/figure&&p&也是山塘街上的一家老店，早上的油氽团子应该算是苏州绝无仅有的，而且只有早上做，卖完为止，皮脆肉香，值得体验。还有生煎、馄饨、面、汤包，糕团也做，很多苏州小吃在这里都能吃到。店堂不新，人气却很足，大部分都是老苏州在吃，这就是好味道的证明。&/p&&p&推荐：油氽团子、生煎&/p&&p&来自苏州土生土长的知友 &a class=&member_mention& href=&//www.zhihu.com/people/acaefaaffca6951b88dfff& data-hash=&acaefaaffca6951b88dfff& data-hovercard=&p$b$acaefaaffca6951b88dfff&&@Valen 狐&/a& 补充到： 生煎只吃苏安新村里的哑巴生煎，那家是老店，就算是观前街的分店都不去吃。&/p&&p&地址：山塘街329号（阿坤卤菜对面）&/p&&p&&br&&/p&&p&05 姚记豆浆&/p&&figure&&img src=&https://pic2.zhimg.com/d6fbd33305_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&397& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic2.zhimg.com/d6fbd33305_r.jpg&&&/figure&&p&吃过咸的豆浆吗？咸豆浆是这家的招牌。24小时经营的老店，老板人特别好。不变的装修，不变的味道。想吃早餐标配豆浆油条的话就来这吧，家里懒得烧粥也可以到这里来一趟。从店里吃完早餐走出来会觉得回到了小时候，街道不宽却很热闹。&/p&&p&推荐：豆浆、油条&/p&&p&地址：西中市60号(老阊门城门)&/p&&p&&br&&/p&&p&&b&06 海珍饮食店&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/dcd33dfdb2e4a2bea3c610_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&550& data-rawheight=&413& class=&origin_image zh-lightbox-thumb& width=&550& data-original=&https://pic1.zhimg.com/dcd33dfdb2e4a2bea3c610_r.jpg&&&/figure&&p&附近的居民都推荐的早餐店，每天早上路过都会很挤。最受好评的是他家的小笼，经常要排队才能买到，皮子是半发酵的，汤汁很多，就算打包带回家再热热也依旧能尝到爆满的汁水。大排面、焖肉面也做得不错，很多人都觉得，在这家店能吃到上世纪八十年代的苏州味道。&/p&&p&推荐：小笼包、大排面&/p&&p&地址：新庄新村（近新庄菜市场）&/p&&p&&br&&/p&&p&&b&07 小姜锅贴&/b&&/p&&figure&&img src=&https://pic3.zhimg.com/783c4cec7b2b061ff9cb7ca_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&431& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic3.zhimg.com/783c4cec7b2b061ff9cb7ca_r.jpg&&&/figure&&p&如果有人问苏州哪里的锅贴好，我就推荐这家。新鲜出炉的锅贴香喷喷的，皮有嚼劲，鲜肉的肉很鲜美，牛肉的味道更浓郁，带着点咖喱味。锅贴底口感类似生煎脆脆的，却又没那么厚，皮子上的芝麻香气很足。早饭一客锅贴，够啦！&/p&&p&推荐：鲜肉锅贴、牛肉锅贴&/p&&p&地址：书院巷81号&/p&&p&tips：除了锅贴生煎来苏州也是必吃中的必吃啦。生煎最负盛名的是最早开在苏安新村的&b&哑巴生煎，&/b&现在市区临顿路菉葭巷对面过桥有分店。生煎绝对的够味，周末早起只为美美的吃上它一口。刚出炉的先凉上那么一会，煎皮薄底脆肉香，一口咬下去汤汁满溢，吃下一客不是问题。配牛肉粉丝汤亦或是泡泡小馄饨都不错。&/p&&p&&a class=&member_mention& href=&//www.zhihu.com/people/c48c1d4914bec050c53776& data-hash=&c48c1d4914bec050c53776& data-hovercard=&p$b$c48c1d4914bec050c53776&&@小旅&/a& ：吃生煎，我是去苏安新村的&b&哑巴生煎&/b&，彩香新村的&b&震源生煎&/b&，山塘街的&b&大阿二生煎&/b&。老西白家的生煎口味一般，还没小杨生煎好吃呢。&/p&&p&&br&&/p&&p&&b&08 万福兴&/b&&/p&&figure&&img src=&https://pic3.zhimg.com/292bd6f9e4e_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&401& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic3.zhimg.com/292bd6f9e4e_r.jpg&&&/figure&&p&又是一家老字号，一年四季的糕团都很好吃。不爱吃糕团的（青团是离开苏州后至今还很怀念的一个小甜点，不管爱不爱吃甜的都请一定替我再尝一口 TwT）到店吃馄饨、面、汤团，他家的汤团也是很好吃的。&/p&&p&推荐：青团子、炒肉团子&/p&&p&地址：东中市12号&/p&&p&&br&&/p&&p&&b&09 潘玉麟糖粥&/b&&/p&&figure&&img src=&https://pic2.zhimg.com/06b506ddc1f7f452eacb5_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&390& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic2.zhimg.com/06b506ddc1f7f452eacb5_r.jpg&&&/figure&&p&名牌小摊位，做的是经典的苏州甜点，爱吃甜品的姑娘们可得留着肚子来这里尝试，糖粥是用糯米精心熬制的，还会加入厚厚的红豆沙和香香的桂花，潘玉麟老夫妇俩笑呵呵地为顾客盛粥，让你还没有吃就已经甜滋滋了。&/p&&p&推荐：桂花酒酿圆子、酒酿小圆子、赤豆小圆子&/p&&p&地址：皮市街172号花鸟市场北门外（史家巷西口）&/p&&p&&br&&/p&&p&&u&&b&（二）平江路必吃美食：&/b&&/u&&/p&&p&&br&&/p&&p&&b&01 鸡脚旮旯&/b&&/p&&figure&&img src=&https://pic3.zhimg.com/a2f17b393ef2_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&581& data-rawheight=&392& class=&origin_image zh-lightbox-thumb& width=&581& data-original=&https://pic3.zhimg.com/a2f17b393ef2_r.jpg&&&/figure&&p&&br&&/p&&p&平江路上有很多卖鸡爪的小店,鸡脚旮旯这家店在网上是一家很红火的店，来平江路当然不能错过。这家店是开在平江路上的一个小巷子里的，门口有大大的旗子招牌，所以很好找。店小小的，门口有人排队，但是流水操作其实还是蛮快，他家把鸡爪，鸭胗，事先装好在牛皮纸里，还配有一次性手套。鸡爪10元4个，鸭胗10元2个，鸡脚煮的很透很软烂，啃起来也方便。鸭胗很入味，而且卤的味道也好，吃起来有嚼劲刚刚好，不会难嚼也不会煮太烂。&/p&&p&推荐：鸡脚、鸭胗&/p&&p&地址：平江路中段邾长巷&/p&&p&tips：有不少人喷这家鸡脚，只能说是平江路那么多家我最推荐的一家。其实苏州还有家最正宗最有味的鸡爪店，叫【&b&祥鑫饮食店&/b&】。位于锦帆路通关坊对面皇废基，距离平江路不会太远。平江路的鸡爪店虽多，但这家就是后来平江路纷纷效仿鸡爪的那家店。鸡爪最出名，脂膏满嘴，久久回味。此外赤豆糊圆子煮的比很多大饭店好，来这里也必点。&/p&&p&&br&&/p&&p&&b&02 苏妃奶酪&/b&&/p&&figure&&img src=&https://pic3.zhimg.com/4f25fc580a488c89bafaeca_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&580& data-rawheight=&397& class=&origin_image zh-lightbox-thumb& width=&580& data-original=&https://pic3.zhimg.com/4f25fc580a488c89bafaeca_r.jpg&&&/figure&&p&苏妃奶酪也是我来平江路必点的小吃之一，通常奶酪+鸡脚是标配。好吃得不要不要的。奶酪口味也很多可以选择，不用担心单一。平江路的各种手工酸奶酪甜点什么的作歇脚之用挺好，但并不属于姑苏美食噢~&/p&&p&推荐：原味奶酪&/p&&p&地址：南石子街(青石桥西侧)&/p&&p&&br&&/p&&p&&b&03 姑苏特色小吃店&/b&&/p&&figure&&img src=&https://pic3.zhimg.com/c2d6ca5b632a_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&403& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic3.zhimg.com/c2d6ca5b632a_r.jpg&&&/figure&&p&说起蟹壳黄又是一阵咽口水，苏州的这家蟹壳黄和扬州东关街的一家蟹壳黄都是心头肉。蟹壳黄其实就是是一种蟹壳大小的小烧饼，咬起来口感有点类似福建的光饼。蟹壳黄一定要趁热吃，酥的满嘴屑。&/p&&p&推荐：蟹壳黄&/p&&p&地址：大儒巷53号（近平江路）&/p&&p&&br&&/p&&p&&b&04 鱼食饭稻土灶馆&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/e2e20c673cf4a2c0bd47ce96a1191724_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&391& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic1.zhimg.com/e2e20c673cf4a2c0bd47ce96a1191724_r.jpg&&&/figure&&p&这是一家吃农家菜的饭馆，装修较为传统，进门处的古装模型、窗边的鸟笼、墙上的汉服劳作图，配合着店内的方桌木椅，使这家店多了一丝古韵。菜单在墙上，带有一丝乡土风情。菜都符合大众口味，分量足，性价比挺高的。人很多，不想排队的话还是要早点去哦。&/p&&p&推荐：鸡头米炒百合 , 剁椒鱼头 , 豆腐 , 椒麻鸡 &/p&&p&地址：平江路68-76号&/p&&p&&br&&/p&&p&&br&&/p&&p&&b&05 随柳居&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/76a2f67a604d1a861d468ea95dcc0fcc_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&838& data-rawheight=&600& class=&origin_image zh-lightbox-thumb& width=&838& data-original=&https://pic1.zhimg.com/76a2f67a604d1a861d468ea95dcc0fcc_r.jpg&&&/figure&&p&简介：主营一些苏州传统特色小吃，开在临顿路的和基广场后面麦当劳那一排，还有其他分店。里头什么酸梅汤赤豆糊圆子冰冻牛奶酒酿各类传统甜品很丰富，填肚子的面跟盖浇饭也有。夏天有苏式凉拌面。桂花糖芋艿有淡淡的桂花清香，汤头浓郁，芋艿又甜又糯，实实在在老苏州的味道。除了苏州小吃甜品，还有苏式汤面、盖浇饭，吃个简餐也是非常不错的选择。&/p&&p&推荐：桂花糖粥、荠菜馄饨、桂花赤豆糊、桂花糖芋艿&/p&&p&地址：建新巷17号&/p&&p&&br&&/p&&p&&u&&b&（三）其他美食：&/b&&/u&&/p&&p&&br&&/p&&p&&br&&/p&&p&&b&01 同德兴面馆&/b&&/p&&figure&&img src=&https://pic4.zhimg.com/46cd39705eea1a448433_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&522& data-rawheight=&376& class=&origin_image zh-lightbox-thumb& width=&522& data-original=&https://pic4.zhimg.com/46cd39705eea1a448433_r.jpg&&&/figure&&p&&br&&/p&&p&这家面馆我认为是必吃中的必吃，游客推荐去十全街那家总店，环境最好。以前吃面口味很重，这种清一色的面我是拒绝吃的。但这家的面味道鲜美到原味就很好入口，再点上几个小菜，热热汤面大口下肚的感觉真的活似神仙。&/p&&p&推荐：红汤/白汤面&/p&&p&地址：十全街624号(近凤凰街)&/p&&p&感谢 &a class=&member_mention& href=&//www.zhihu.com/people/b48521fcf952a5b66e0f3c& data-hash=&b48521fcf952a5b66e0f3c& data-hovercard=&p$b$b48521fcf952a5b66e0f3c&&@Ayrinnnnn&/a& 提醒：同得兴十全街店13:00就打烊，07:00-13:00营业，大家不要想吃的话只能选择早上的时间啦。&/p&&p&tips：关于苏式面，可选性很大，口味这东西向来见仁见智，我个人最爱同德兴的。想要看详细完整的面测评和推荐，可以看看这个专栏的文章，写得不错&a href=&https://zhuanlan.zhihu.com/p/& class=&internal&&要说吃面，还是苏州的细腻 - 旅人說 - 知乎专栏&/a&&/p&&p&&br&&/p&&p&&b&02 美味渔家蟹园 &/b&&/p&&figure&&img src=&https://pic1.zhimg.com/17ad2923abfd972c5e8c5d4_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&407& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic1.zhimg.com/17ad2923abfd972c5e8c5d4_r.jpg&&&/figure&&p&&br&&/p&&p&太湖的一家农家乐，来这里吃“太湖三白”很推荐，味道正宗、价格合理。老板服务很周到，会帮顾客挑选肥壮的大闸蟹，现捞现煮，一切透明化，让你吃的放心。土鸡也是自家养殖，太湖三白选料讲究，农家乐风情呈现，有着朴实的幸福感。来太湖除了吃知名的太湖三白，另外再推荐个白鳗，清蒸黄焖都好吃，跟豆腐似的入口即化。&/p&&p&推荐：太湖三白、土鸡汤&/p&&p&地址：吴中区金庭镇金铎村白塔湾（太湖旁）&/p&&p&&br&&/p&&p&&b&03 藏书羊肉&/b&&/p&&figure&&img src=&https://pic4.zhimg.com/61a8abff8d3123_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&417& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic4.zhimg.com/61a8abff8d3123_r.jpg&&&/figure&&p&藏书羊肉是江苏省苏州汉族风味小吃。历史悠久，以其独特的烧煮技艺，烹调成各式羊肉菜肴。烧煮而成，白烧以汤色乳白，香气浓郁，肉酥而不烂，口感鲜而不腻。&/p&&p&推荐：羊肉火锅&/p&&p&地址：平江区临顿路204号(全大鱼羊馆旁边)&/p&&p&tips：最最最正宗的臧书羊肉当然还得去&b&藏书镇&/b&品尝了，藏书镇的协会评定了羊肉餐馆的“星级”，门口有铜牌，其中最高“3星级”的大约有7-8家，价格都差不多，比非3星级的稍贵，但七人200元左右已经吃得很好了。推荐店家：&b&老庆泰&/b&，位置在藏书镇最繁华的地段，藏书羊肉美食城门大招牌下第一家。&/p&&p&另外姑苏区定慧寺巷里的&b&兴奋羊肉店&/b&也值得一试，每人15块钱左右的羊肉羊肚加白菜粉丝血油豆腐做个明炉再切点羊杂，再来俩小菜真是湿冷的苏州寒夜里最大的享受了。&/p&&p&&br&&/p&&p&&b&04 绿杨馄饨&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/19ebd27eaaa_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&389& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic1.zhimg.com/19ebd27eaaa_r.jpg&&&/figure&&p&&br&&/p&&p&又一个苏州百年老字号的店，分店有很多家。推荐的这家位于临顿路。他们家的汤是用老母鸡熬制的，所以味道特别鲜美，而且馄饨是当天包的，不用冷冻的馄饨，所以吃起来特别好吃。绿杨的馄饨皮薄馅儿多，汤汁清亮，肉质也很有弹性。&/p&&p&推荐：三鲜馄饨&/p&&p&地址：临顿路239号&/p&&p&tips：如果要吃绿杨，请吃&b&临顿路&/b&的那家，观前街的不建议去，游客多的地方食物的原汁原味要打个问号。&/p&&p&苏州馄饨也是必尝的一个小吃，除了知名的绿杨馄饨外，还有几家不错的：&/p&&p&&b&乐惠馄饨店&/b&，坐标吴趋坊44号。刘嘉玲说她小时候住那边从小吃这家，他家馄饨汤团的皮吃起来有别于以往吃过的。泡泡馄饨最为推荐。他家的汤团也很棒，皮子软糯有弹性但不黏牙馅很扎实。&/p&&p&&b&味雅馄饨店&/b&，坐标锦帆路通关坊对面皇废基斜坡上（乐桥的东南方向）。特色是骨头汤底以及口味丰富的大馄饨，小馄饨也不错。大馄饨有芹菜肉韭菜肉荠菜肉等。馄饨都是现包的，很新鲜。&/p&&p&&br&&/p&&p&&b&05 得月楼&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/d135af780cc_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&570& data-rawheight=&380& class=&origin_image zh-lightbox-thumb& width=&570& data-original=&https://pic1.zhimg.com/d135af780cc_r.jpg&&&/figure&&p&&br&&/p&&p&如果说去杭州一定要吃的正餐是楼外楼，那么来苏州必定是得月楼了。观前街附近的太监弄里的一家百年老店，圆形拱门、高高悬挂的灯笼、木质大圆桌，古色古香，颇有味道，慕名而来的食客络绎不绝。上菜速度很快，虽然有点小贵，但是味道真的是没话说的。既然来了苏州，那么作为苏州最有名的饭店之一，你一定要来尝试一下。&/p&&p&推荐：荷叶粉蒸肉 , 松鼠桂鱼 , 碧螺虾仁 , 樱桃肉&/p&&p&地址：太监弄43号&/p&&p&&a class=&member_mention& href=&//www.zhihu.com/people/4fd23cc3e90& data-hash=&4fd23cc3e90& data-hovercard=&p$b$4fd23cc3e90&&@曹茗然&/a& 补充到：近些年松鹤楼得月楼的苏帮菜水准都下降得厉害，作为土生土长的苏州人，强推凤凰街上的&b&协和菜馆&/b&，他家的味道更贴近苏帮的原汁原味 。&/p&&p&tips：协和菜馆是老字号的苏帮菜馆，店面不大。网友称这个饭馆是平民的价格，松鹤楼的味道。龙井虾仁和松鼠鳜鱼必点，是招牌菜。tips：协和菜馆是老字号的苏帮菜馆，店面不大。网友称这个饭馆是平民的价格，松鹤楼的味道。龙井虾仁和松鼠鳜鱼必点，是招牌菜。&/p&&p&&br&&/p&&p&&b&06 黄天源&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/a457a57b6b36decd7fca6f_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&600& data-rawheight=&398& class=&origin_image zh-lightbox-thumb& width=&600& data-original=&https://pic1.zhimg.com/a457a57b6b36decd7fca6f_r.jpg&&&/figure&&p&苏州玄妙观隔壁的黄天源，是家响当当的糕团老字号了。过去，苏州人大都爱吃甜糯之食，糕团食品自古也是苏州人时令饮食的一种习俗。窗口有散装的现制点心，店内有包装的特产手信。虽然糕团是这家店的一段佳话，但答主还是最喜欢青团呐~&/p&&p&推荐：糕团、薄荷糕、玫瑰松饼必吃；黄天源的面，也属于苏州十碗面之一，值得一试。&/p&&p&地址：观前街86号&/p&&p&&br&&/p&&p&&b&07 湘城人和园&/b&&/p&&figure&&img src=&https://pic4.zhimg.com/d01ed4230f_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&700& data-rawheight=&467& class=&origin_image zh-lightbox-thumb& width=&700& data-original=&https://pic4.zhimg.com/d01ed4230f_r.jpg&&&/figure&&p&简介：他们家有着最苏州味儿的麻饼，老字号，枣泥馅的里面还有松子，软软的，甜而不腻。特色在于里头有猪油，猪油融化了会让枣泥不那么干，香味更浓郁。非常建议买回去作为伴手礼。私信也会收到一些小伙伴问去苏州带什么伴手礼特产回家，推荐这家的麻饼和黄天源的糕团。建议在微波炉加热30秒，不是大热天放一个月没问题。&/p&&p&推荐：麻饼&/p&&p&地址：王天井巷61号（分店有好几家，大家可以地图搜一下就近前往）&/p&&p&&br&&/p&&p&&b&08 石家饭店&/b&&/p&&figure&&img src=&https://pic2.zhimg.com/8b2fc0fff091eb6a25aa497f88bd6345_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&700& data-rawheight=&525& class=&origin_image zh-lightbox-thumb& width=&700& data-original=&https://pic2.zhimg.com/8b2fc0fff091eb6a25aa497f88bd6345_r.jpg&&&/figure&&p&简介：如果你来&b&木渎&/b&或者&b&李公堤&/b&游玩的话，可以预留时间来石家饭店吃吃苏帮菜，也是老字号菜馆。石家饭店因乾隆下江南时曾在这里吃饭而著名，据传乾隆在这里吃了鲃肺汤和活炝河虾。于右任先生有诗：“老桂开花天下香，看花走遍太湖旁，归舟木渎犹堪记，多谢石家鲃肺汤”。赞扬了这里的鲃肺汤。&/p&&p&推荐：酱方、清炒河虾仁、三虾豆腐、鲃肺汤&/p&&p&地址：木渎古镇中市街18号；李公堤三期三号楼(近机场路)&/p&&p&tips：苏帮菜除了石家饭店，&b&吴门人家&/b&和&b&珍珠饭店&/b&也不错。&/p&&p&&br&&/p&&p&&b&09 长发肉月饼&/b&&/p&&figure&&img src=&https://pic1.zhimg.com/ebb80e6dd1c339a4b9240_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&528& data-rawheight=&373& class=&origin_image zh-lightbox-thumb& width=&528& data-original=&https://pic1.zhimg.com/ebb80e6dd1c339a4b9240_r.jpg&&&/figure&&p&苏式月饼皮，里面是鲜肉，有汤汁，因此只能现吃（也是甜咸口的肉）。临顿路的长发西饼口买月饼的从早上七点多就开始排队，而且前面常有人四五十个一买，更有甚者一口气买一百个。这夸张的排队有力地证实了肉月饼的吸引力和馋嘴度。看看一锅锅出炉的月饼，看看一张张快点买到月饼期盼的脸，这就是它的魅力，无可替代。除了以往的鲜肉月饼和虾肉月饼之外，这几年更是推出了肉松月饼和榴莲月饼。虽然推出些其他口味，但鲜肉却是永远的经典。&/p&&p&&br&&/p&&p&&br&&/p&&p&以上图文部分摘自网络，侵删。&/p&
破8千赞，回答上了知乎日报啦～已删去鸡肋推荐并增加了更为地道的美食，剩下的都是集锦。大学四年在苏州，谈一谈眼里的苏州好去处和好味道。 旅行必体验之景点篇： 1 苏州大学看老式红楼和钟楼博物馆和大学是我认为一个城市的精髓所在之一，苏州大学本部就…
&figure&&img src=&https://pic4.zhimg.com/v2-a70a0eeb6dc09d86c38b1b2e9dc129a6_b.jpg& data-rawwidth=&609& data-rawheight=&434& class=&origin_image zh-lightbox-thumb& width=&609& data-original=&https://pic4.zhimg.com/v2-a70a0eeb6dc09d86c38b1b2e9dc129a6_r.jpg&&&/figure&&p&最近，微信小游戏跳一跳可以说是火遍了全国，从小孩子到大孩子仿佛每一个人都在刷跳一跳，作为无（zhi）所（hui）不（ban）能（zhuan）的AI程序员，我们在想，能不能用人工智能（AI）和计算机视觉（CV）的方法来玩一玩这个游戏？于是，我们开发了微信跳一跳Auto-Jump算法，重新定义了玩跳一跳的正确姿势，我们的算法不仅远远超越了人类的水平，在速度和准确度上也远远超越了目前已知的所有算法，可以说是跳一跳界的state-of-the-art，下面我们详细介绍我们的算法。&/p&&p&&br&&/p&&p&算法的第一步是获取手机屏幕的截图并可以控制手机的触控操作，我们的github仓库里详细介绍了针对Android和IOS手机的配置方法，&/p&&a href=&http://link.zhihu.com/?target=https%3A//github.com/Prinsphield/Wechat_AutoJump& data-draft-node=&block& data-draft-type=&link-card& data-image=&https://pic1.zhimg.com/v2-db086a5e6c_ipico.jpg& data-image-width=&420& data-image-height=&420& class=& wrap external& target=&_blank& rel=&nofollow noreferrer&&Prinsphield/Wechat_AutoJump&/a&&p&你只需要按照将手机连接电脑，按照教程执行就可以完成配置。在获取到屏幕截图之后，就是个简单的视觉问题。我们需要找的就是小人的位置和下一次需要跳的台面的中心。&/p&&p&如图所示，绿色的点代表小人当前的位置，红点代表目标位置。&/p&&figure&&img src=&https://pic2.zhimg.com/v2-b316c887ce6ac13198a29_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&1060& data-rawheight=&936& class=&origin_image zh-lightbox-thumb& width=&1060& data-original=&https://pic2.zhimg.com/v2-b316c887ce6ac13198a29_r.jpg&&&/figure&&h2&多尺度搜索（Multiscale Search）&/h2&&p&这个问题可以有非常多的方法去解，为了糙快猛地刷上榜，我一开始用的方式是多尺度搜索。我随便找了一张图，把小人抠出来，就像下面这样。&/p&&figure&&img src=&https://pic2.zhimg.com/v2-d89c675ab59f1_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&50& data-rawheight=&140& class=&content_image& width=&50&&&/figure&&p&另外，我注意到小人在屏幕的不同位置，大小略有不同，所以我设计了多尺度的搜索，用不同大小的进行匹配，最后选取置信度（confidence score）最高的。&/p&&p&多尺度搜索的代码长这样&/p&&div class=&highlight&&&pre&&code class=&language-python3&&&span&&/span&&span class=&k&&def&/span& &span class=&nf&&multi_scale_search&/span&&span class=&p&&(&/span&&span class=&n&&pivot&/span&&span class=&p&&,&/span& &span class=&n&&screen&/span&&span class=&p&&,&/span& &span class=&nb&&range&/span&&span class=&o&&=&/span&&span class=&mf&&0.3&/span&&span class=&p&&,&/span& &span class=&n&&num&/span&&span class=&o&&=&/span&&span class=&mi&&10&/span&&span class=&p&&):&/span&
&span class=&n&&H&/span&&span class=&p&&,&/span& &span class=&n&&W&/span& &span class=&o&&=&/span& &span class=&n&&screen&/span&&span class=&o&&.&/span&&span class=&n&&shape&/span&&span class=&p&&[:&/span&&span class=&mi&&2&/span&&span class=&p&&]&/span&
&span class=&n&&h&/span&&span class=&p&&,&/span& &span class=&n&&w&/span& &span class=&o&&=&/span& &span class=&n&&pivot&/span&&span class=&o&&.&/span&&span class=&n&&shape&/span&&span class=&p&&[:&/span&&span class=&mi&&2&/span&&span class=&p&&]&/span&
&span class=&n&&found&/span& &span class=&o&&=&/span& &span class=&kc&&None&/span&
&span class=&k&&for&/span& &span class=&n&&scale&/span& &span class=&ow&&in&/span& &span class=&n&&np&/span&&span class=&o&&.&/span&&span class=&n&&linspace&/span&&span class=&p&&(&/span&&span class=&mi&&1&/span&&span class=&o&&-&/span&&span class=&nb&&range&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&o&&+&/span&&span class=&nb&&range&/span&&span class=&p&&,&/span& &span class=&n&&num&/span&&span class=&p&&)[::&/span&&span class=&o&&-&/span&&span class=&mi&&1&/span&&span class=&p&&]:&/span&
&span class=&n&&resized&/span& &span class=&o&&=&/span& &span class=&n&&cv2&/span&&span class=&o&&.&/span&&span class=&n&&resize&/span&&span class=&p&&(&/span&&span class=&n&&screen&/span&&span class=&p&&,&/span& &span class=&p&&(&/span&&span class=&nb&&int&/span&&span class=&p&&(&/span&&span class=&n&&W&/span& &span class=&o&&*&/span& &span class=&n&&scale&/span&&span class=&p&&),&/span& &span class=&nb&&int&/span&&span class=&p&&(&/span&&span class=&n&&H&/span& &span class=&o&&*&/span& &span class=&n&&scale&/span&&span class=&p&&)))&/span&
&span class=&n&&r&/span& &span class=&o&&=&/span& &span class=&n&&W&/span& &span class=&o&&/&/span& &span class=&nb&&float&/span&&span class=&p&&(&/span&&span class=&n&&resized&/span&&span class=&o&&.&/span&&span class=&n&&shape&/span&&span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&])&/span&
&span class=&k&&if&/span& &span class=&n&&resized&/span&&span class=&o&&.&/span&&span class=&n&&shape&/span&&span class=&p&&[&/span&&span class=&mi&&0&/span&&span class=&p&&]&/span& &span class=&o&&&&/span& &span class=&n&&h&/span& &span class=&ow&&or&/span& &span class=&n&&resized&/span&&span class=&o&&.&/span&&span class=&n&&shape&/span&&span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&]&/span& &span class=&o&&&&/span& &span class=&n&&w&/span&&span class=&p&&:&/span&
&span class=&k&&break&/span&
&span class=&n&&res&/span& &span class=&o&&=&/span& &span class=&n&&cv2&/span&&span class=&o&&.&/span&&span class=&n&&matchTemplate&/span&&span class=&p&&(&/span&&span class=&n&&resized&/span&&span class=&p&&,&/span& &span class=&n&&pivot&/span&&span class=&p&&,&/span& &span class=&n&&cv2&/span&&span class=&o&&.&/span&&span class=&n&&TM_CCOEFF_NORMED&/span&&span class=&p&&)&/span&
&span class=&n&&loc&/span& &span class=&o&&=&/span& &span class=&n&&np&/span&&span class=&o&&.&/span&&span class=&n&&where&/span&&span class=&p&&(&/span&&span class=&n&&res&/span& &span class=&o&&&=&/span& &span class=&n&&res&/span&&span class=&o&&.&/span&&span class=&n&&max&/span&&span class=&p&&())&/span&
&span class=&n&&pos_h&/span&&span class=&p&&,&/span& &span class=&n&&pos_w&/span& &span class=&o&&=&/span& &span class=&nb&&list&/span&&span class=&p&&(&/span&&span class=&nb&&zip&/span&&span class=&p&&(&/span&&span class=&o&&*&/span&&span class=&n&&loc&/span&&span class=&p&&))[&/span&&span class=&mi&&0&/span&&span class=&p&&]&/span&
&span class=&k&&if&/span& &span class=&n&&found&/span& &span class=&ow&&is&/span& &span class=&kc&&None&/span& &span class=&ow&&or&/span& &span class=&n&&res&/span&&span class=&o&&.&/span&&span class=&n&&max&/span&&span class=&p&&()&/span& &span class=&o&&&&/span& &span class=&n&&found&/span&&span class=&p&&[&/span&&span class=&o&&-&/span&&span class=&mi&&1&/span&&span class=&p&&]:&/span&
&span class=&n&&found&/span& &span class=&o&&=&/span& &span class=&p&&(&/span&&span class=&n&&pos_h&/span&&span class=&p&&,&/span& &span class=&n&&pos_w&/span&&span class=&p&&,&/span& &span class=&n&&r&/span&&span class=&p&&,&/span& &span class=&n&&res&/span&&span class=&o&&.&/span&&span class=&n&&max&/span&&span class=&p&&())&/span&
&span class=&k&&if&/span& &span class=&n&&found&/span& &span class=&ow&&is&/span& &span class=&kc&&None&/span&&span class=&p&&:&/span& &span class=&k&&return&/span& &span class=&p&&(&/span&&span class=&mi&&0&/span&&span class=&p&&,&/span&&span class=&mi&&0&/span&&span class=&p&&,&/span&&span class=&mi&&0&/span&&span class=&p&&,&/span&&span class=&mi&&0&/span&&span class=&p&&,&/span&&span class=&mi&&0&/span&&span class=&p&&)&/span&
&span class=&n&&pos_h&/span&&span class=&p&&,&/span& &span class=&n&&pos_w&/span&&span class=&p&&,&/span& &span class=&n&&r&/span&&span class=&p&&,&/span& &span class=&n&&score&/span& &span class=&o&&=&/span& &span class=&n&&found&/span&
&span class=&n&&start_h&/span&&span class=&p&&,&/span& &span class=&n&&start_w&/span& &span class=&o&&=&/span& &span class=&nb&&int&/span&&span class=&p&&(&/span&&span class=&n&&pos_h&/span& &span class=&o&&*&/span& &span class=&n&&r&/span&&span class=&p&&),&/span& &span class=&nb&&int&/span&&span class=&p&&(&/span&&span class=&n&&pos_w&/span& &span class=&o&&*&/span& &span class=&n&&r&/span&&span class=&p&&)&/span&
&span class=&n&&end_h&/span&&span class=&p&&,&/span& &span class=&n&&end_w&/span& &span class=&o&&=&/span& &span class=&nb&&int&/span&&span class=&p&&((&/span&&span class=&n&&pos_h&/span& &span class=&o&&+&/span& &span class=&n&&h&/span&&span class=&p&&)&/span& &span class=&o&&*&/span& &span class=&n&&r&/span&&span class=&p&&),&/span& &span class=&nb&&int&/span&&span class=&p&&((&/span&&span class=&n&&pos_w&/span& &span class=&o&&+&/span& &span class=&n&&w&/span&&span class=&p&&)&/span& &span class=&o&&*&/span& &span class=&n&&r&/span&&span class=&p&&)&/span&
&span class=&k&&return&/span& &span class=&p&&[&/span&&span class=&n&&start_h&/span&&span class=&p&&,&/span& &span class=&n&&start_w&/span&&span class=&p&&,&/span& &span class=&n&&end_h&/span&&span class=&p&&,&/span& &span class=&n&&end_w&/span&&span class=&p&&,&/span& &span class=&n&&score&/span&&span class=&p&&]&/span&
&/code&&/pre&&/div&&p&&br&&/p&&p&我们来试一试，效果还不错，应该说是又快又好，我所有的实验中找小人从来没有失误。&/p&&p&不过这里的位置框的底部中心并不是小人的位置，真实的位置是在那之上一些。&/p&&figure&&img src=&https://pic2.zhimg.com/v2-a2d6ecfff13eadf45bff_b.jpg& data-caption=&& data-size=&small& data-rawwidth=&720& data-rawheight=&1280& class=&origin_image zh-lightbox-thumb& width=&720& data-original=&https://pic2.zhimg.com/v2-a2d6ecfff13eadf45bff_r.jpg&&&/figure&&p&同理，目标台面也可以用这种办法搜索，但是我们需要收集一些不同的台面，有圆形的，方形的，便利店，井盖，棱柱等等。由于数量一多，加上多尺度的原因，速度上会慢下来。这时候，我们就需要想办法加速了。首先可以注意到目标位置始终在小人的位置的上面，所以可以操作的一点就是在找到小人位置之后把小人位置以下的部分都舍弃掉，这样可以减少搜索空间。但是这还是不够，我们需要进一步去挖掘游戏里的故事。小人和目标台面基本上是关于屏幕中心对称的位置的。这提供了一个非常好的思路去缩小搜索空间。假设屏幕分辨率是（）的，小人底部的位置是（h1, w1),那么关于中心对称点的位置就是（1280 - h1， 720 - w1），以这个点为中心的一个边长300的正方形内，我们再去多尺度搜索目标位置，就会又快有准了。效果见下图，蓝色框是（300，300）的搜索区域，红色框是搜到的台面，矩形中心就是目标点的坐标了。&/p&&figure&&img src=&https://pic1.zhimg.com/v2-33903afaaee9a8a12a804_b.jpg& data-caption=&& data-size=&small& data-rawwidth=&720& data-rawheight=&1280& class=&origin_image zh-lightbox-thumb& width=&720& data-original=&https://pic1.zhimg.com/v2-33903afaaee9a8a12a804_r.jpg&&&/figure&&h2&加速的奇技淫巧（Fast-Search）&/h2&&p&玩游戏需要细心观察。我们可以发现，小人上一次如果跳到台面中心，那么下一次目标台面的中心会有一个白点，就像刚才所展示的图里的。更加细心的人会发现，白点的RGB值是（245，245，245），这就让我找到了一个非常简单并且高效的方式，就是直接去搜索这个白点，注意到白点是一个连通区域，像素值为（245，245，245）的像素个数稳定在280-310之间，所以我们可以利用这个去直接找到目标的位置。这种方式只在前一次跳到中心的时候可以用，不过没有关系，我们每次都可以试一试这个不花时间的方法，不行再考虑多尺度搜索。&/p&&p&&br&&/p&&p&讲到这里，我们的方法已经可以运行的非常出色了，基本上是一个永动机。下面是用我的手机玩了一个半小时左右，跳了859次的状态，我们的方法正确的计算出来了小人的位置和目标位置，不过我选择狗带了，因为手机卡的已经不行了。&/p&&figure&&img src=&https://pic4.zhimg.com/v2-010cfebde9f1dd02be5ebe100ee507a3_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&1572& data-rawheight=&940& class=&origin_image zh-lightbox-thumb& width=&1572& data-original=&https://pic4.zhimg.com/v2-010cfebde9f1dd02be5ebe100ee507a3_r.jpg&&&/figure&&p&这里有一个示例视频，欢迎观看!&/p&&a class=&video-box& href=&http://link.zhihu.com/?target=https%3A//www.zhihu.com/video/309056& target=&_blank& data-video-id=&& data-video-playable=&true& data-name=&& data-poster=&https://pic1.zhimg.com/80/v2-66e68c0c366ffeebdbfa5ce8_b.jpg& data-lens-id=&309056&&
&img class=&thumbnail& src=&https://pic1.zhimg.com/80/v2-66e68c0c366ffeebdbfa5ce8_b.jpg&&&span class=&content&&
&span class=&title&&&span class=&z-ico-extern-gray&&&/span&&span class=&z-ico-extern-blue&&&/span&&/span&
&span class=&url&&&span class=&z-ico-video&&&/span&https://www.zhihu.com/video/309056&/span&
&p&&br&&/p&&p&&b&到这里就结束了吗？那我们和业余玩家有什么区别？下面进入正经的学术时间，非战斗人员请迅速撤离！&/b&&/p&&p&&br&&/p&&h2&CNN Coarse-to-Fine 模型&/h2&&p&考虑到IOS设备由于屏幕抓取方案的限制（WebDriverAgent获得的截图经过了压缩，图像像素受损，不再是原来的像素值，原因不详，欢迎了解详情的小伙伴提出改进意见~）无法使用fast-search，同时为了兼容多分辨率设备，我们使用卷积神经网络构建了一个更快更鲁棒的目标检测模型，下面分数据采集与预处理，coarse模型，fine模型，cascade四部分介绍我们的算法。&/p&&p&&br&&/p&&h2&数据采集与预处理&/h2&&p&基于我们非常准确的multiscale-search、fast-search模型，我们采集了7次实验数据，共计大约3000张屏幕截图，每一张截图均带有目标位置标注，对于每一张图，我们进行了两种不同的预处理方式，并分别用于训练coarse模型和fine模型，下面分别介绍两种不同的预处理方式。&/p&&p&&br&&/p&&h2&Coarse 模型数据预处理&/h2&&p&由于每一张图像中真正对于当前判断有意义的区域只在屏幕中央位置，即人和目标物体所在的位置，因此，每一张截图的上下两部分都是没有意义的，因此，我们将采集到的大小为的图像沿x方向上下各截去320*720大小，只保留中心640*720的图像作为训练数据。&/p&&p&&br&&/p&&p&我们观察到，游戏中，每一次当小人落在目标物中心位置时，下一个目标物的中心会出现一个白色的圆点，&/p&&figure&&img src=&https://pic4.zhimg.com/v2-cfc58b47f323_b.jpg& data-caption=&& data-size=&small& data-rawwidth=&720& data-rawheight=&1280& class=&origin_image zh-lightbox-thumb& width=&720& data-original=&https://pic4.zhimg.com/v2-cfc58b47f323_r.jpg&&&/figure&&p&考虑到训练数据中fast-search会产生大量有白点的数据，为了杜绝白色圆点对网络训练的干扰，我们对每一张图进行了去白点操作，具体做法是，用白点周围的纯色像素填充白点区域。&/p&&p&&br&&/p&&h2&Fine 模型数据预处理&/h2&&p&为了进一步提升模型的精度，我们为fine模型建立了数据集，对训练集中的每一张图，在目标点附近截取320*320大小的一块作为训练数据，&/p&&figure&&img src=&https://pic1.zhimg.com/v2-d75ceb625db7b9e95c9dc50_b.jpg& data-caption=&& data-size=&normal& data-rawwidth=&320& data-rawheight=&320& class=&content_image& width=&320&&&/figure&&p&为了防止网络学到trivial的结果，我们对每一张图增加了50像素的随机偏移。fine模型数据同样进行了去白点操作。&/p&&p&&br&&/p&&h2&Coarse 模型&/h2&&p&我们把这一问题看成了回归问题，coarse模型使用一个卷积神经网络回归目标的位置，&/p&&div class=&highlight&&&pre&&code class=&language-python3&&&span&&/span&&span class=&k&&def&/span& &span class=&nf&&forward&/span&&span class=&p&&(&/span&&span class=&bp&&self&/span&&span class=&p&&,&/span& &span class=&n&&img&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&,&/span& &span class=&n&&keep_prob&/span&&span class=&p&&,&/span& &span class=&n&&name&/span&&span class=&o&&=&/span&&span class=&s1&&'coarse'&/span&&span class=&p&&):&/span&
&span class=&k&&with&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&name_scope&/span&&span class=&p&&(&/span&&span class=&n&&name&/span&&span class=&p&&):&/span&
&span class=&k&&with&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&variable_scope&/span&&span class=&p&&(&/span&&span class=&n&&name&/span&&span class=&p&&):&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&conv2d&/span&&span class=&p&&(&/span&&span class=&s1&&'conv1'&/span&&span class=&p&&,&/span& &span class=&n&&img&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&3&/span&&span class=&p&&,&/span& &span class=&mi&&3&/span&&span class=&p&&,&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&input_channle&/span&&span class=&p&&,&/span& &span class=&mi&&16&/span&&span class=&p&&],&/span& &span class=&mi&&2&/span&&span class=&p&&)&/span&
&span class=&c1&&# out = tf.layers.batch_normalization(out, name='bn1', training=is_training)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&relu&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&n&&name&/span&&span class=&o&&=&/span&&span class=&s1&&'relu1'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_conv_bn_relu&/span&&span class=&p&&(&/span&&span class=&s1&&'conv2'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&3&/span&&span class=&p&&,&/span& &span class=&mi&&3&/span&&span class=&p&&,&/span& &span class=&mi&&16&/span&&span class=&p&&,&/span& &span class=&mi&&32&/span&&span class=&p&&],&/span& &span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&max_pool&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&n&&padding&/span&&span class=&o&&=&/span&&span class=&s1&&'SAME'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_conv_bn_relu&/span&&span class=&p&&(&/span&&span class=&s1&&'conv3'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&5&/span&&span class=&p&&,&/span& &span class=&mi&&5&/span&&span class=&p&&,&/span& &span class=&mi&&32&/span&&span class=&p&&,&/span& &span class=&mi&&64&/span&&span class=&p&&],&/span& &span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&max_pool&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&n&&padding&/span&&span class=&o&&=&/span&&span class=&s1&&'SAME'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_conv_bn_relu&/span&&span class=&p&&(&/span&&span class=&s1&&'conv4'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&7&/span&&span class=&p&&,&/span& &span class=&mi&&7&/span&&span class=&p&&,&/span& &span class=&mi&&64&/span&&span class=&p&&,&/span& &span class=&mi&&128&/span&&span class=&p&&],&/span& &span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&max_pool&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&n&&padding&/span&&span class=&o&&=&/span&&span class=&s1&&'SAME'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_conv_bn_relu&/span&&span class=&p&&(&/span&&span class=&s1&&'conv5'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&9&/span&&span class=&p&&,&/span& &span class=&mi&&9&/span&&span class=&p&&,&/span& &span class=&mi&&128&/span&&span class=&p&&,&/span& &span class=&mi&&256&/span&&span class=&p&&],&/span& &span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&max_pool&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&n&&padding&/span&&span class=&o&&=&/span&&span class=&s1&&'SAME'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&reshape&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&o&&-&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&256&/span& &span class=&o&&*&/span& &span class=&mi&&20&/span& &span class=&o&&*&/span& &span class=&mi&&23&/span&&span class=&p&&])&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_fc&/span&&span class=&p&&(&/span&&span class=&s1&&'fc1'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&256&/span& &span class=&o&&*&/span& &span class=&mi&&20&/span& &span class=&o&&*&/span& &span class=&mi&&23&/span&&span class=&p&&,&/span& &span class=&mi&&256&/span&&span class=&p&&],&/span& &span class=&n&&keep_prob&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_fc&/span&&span class=&p&&(&/span&&span class=&s1&&'fc2'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&256&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&],&/span& &span class=&n&&keep_prob&/span&&span class=&p&&)&/span&
&span class=&k&&return&/span& &span class=&n&&out&/span&
&/code&&/pre&&/div&&p&经过十小时的训练，coarse模型在测试集上达到了6像素的精度，实际测试精度大约为10像素，在测试机器（MacBook Pro Retina, 15-inch, Mid
GHz Intel Core i7）上inference时间0.4秒。这一模型可以很轻松的拿到超过1k的分数，这已经远远超过了人类水平和绝大多数自动算法的水平，日常娱乐完全够用，不过，你认为我们就此为止那就大错特错了~&/p&&p&&br&&/p&&h2&Fine 模型&/h2&&p&fine模型结构与coarse模型类似，参数量稍大，fine模型作为对coarse模型的refine操作，&/p&&div class=&highlight&&&pre&&code class=&language-python3&&&span&&/span&&span class=&k&&def&/span& &span class=&nf&&forward&/span&&span class=&p&&(&/span&&span class=&bp&&self&/span&&span class=&p&&,&/span& &span class=&n&&img&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&,&/span& &span class=&n&&keep_prob&/span&&span class=&p&&,&/span& &span class=&n&&name&/span&&span class=&o&&=&/span&&span class=&s1&&'fine'&/span&&span class=&p&&):&/span&
&span class=&k&&with&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&name_scope&/span&&span class=&p&&(&/span&&span class=&n&&name&/span&&span class=&p&&):&/span&
&span class=&k&&with&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&variable_scope&/span&&span class=&p&&(&/span&&span class=&n&&name&/span&&span class=&p&&):&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&conv2d&/span&&span class=&p&&(&/span&&span class=&s1&&'conv1'&/span&&span class=&p&&,&/span& &span class=&n&&img&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&3&/span&&span class=&p&&,&/span& &span class=&mi&&3&/span&&span class=&p&&,&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&input_channle&/span&&span class=&p&&,&/span& &span class=&mi&&16&/span&&span class=&p&&],&/span& &span class=&mi&&2&/span&&span class=&p&&)&/span&
&span class=&c1&&# out = tf.layers.batch_normalization(out, name='bn1', training=is_training)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&relu&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&n&&name&/span&&span class=&o&&=&/span&&span class=&s1&&'relu1'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_conv_bn_relu&/span&&span class=&p&&(&/span&&span class=&s1&&'conv2'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&3&/span&&span class=&p&&,&/span& &span class=&mi&&3&/span&&span class=&p&&,&/span& &span class=&mi&&16&/span&&span class=&p&&,&/span& &span class=&mi&&64&/span&&span class=&p&&],&/span& &span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&max_pool&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&n&&padding&/span&&span class=&o&&=&/span&&span class=&s1&&'SAME'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_conv_bn_relu&/span&&span class=&p&&(&/span&&span class=&s1&&'conv3'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&5&/span&&span class=&p&&,&/span& &span class=&mi&&5&/span&&span class=&p&&,&/span& &span class=&mi&&64&/span&&span class=&p&&,&/span& &span class=&mi&&128&/span&&span class=&p&&],&/span& &span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&max_pool&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&n&&padding&/span&&span class=&o&&=&/span&&span class=&s1&&'SAME'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_conv_bn_relu&/span&&span class=&p&&(&/span&&span class=&s1&&'conv4'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&7&/span&&span class=&p&&,&/span& &span class=&mi&&7&/span&&span class=&p&&,&/span& &span class=&mi&&128&/span&&span class=&p&&,&/span& &span class=&mi&&256&/span&&span class=&p&&],&/span& &span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&max_pool&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&n&&padding&/span&&span class=&o&&=&/span&&span class=&s1&&'SAME'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&bp&&self&/span&&span class=&o&&.&/span&&span class=&n&&make_conv_bn_relu&/span&&span class=&p&&(&/span&&span class=&s1&&'conv5'&/span&&span class=&p&&,&/span& &span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&9&/span&&span class=&p&&,&/span& &span class=&mi&&9&/span&&span class=&p&&,&/span& &span class=&mi&&256&/span&&span class=&p&&,&/span& &span class=&mi&&512&/span&&span class=&p&&],&/span& &span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&n&&is_training&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&nn&/span&&span class=&o&&.&/span&&span class=&n&&max_pool&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&p&&[&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&2&/span&&span class=&p&&,&/span& &span class=&mi&&1&/span&&span class=&p&&],&/span& &span class=&n&&padding&/span&&span class=&o&&=&/span&&span class=&s1&&'SAME'&/span&&span class=&p&&)&/span&
&span class=&n&&out&/span& &span class=&o&&=&/span& &span class=&n&&tf&/span&&span class=&o&&.&/span&&span class=&n&&reshape&/span&&span class=&p&&(&/span&&span class=&n&&out&/span&&span class=&p&&,&/span& &span class=&p&&[&/span&&span class=&o&&-&/span&&span class=&mi&&1&/span&&span class=&p&&,&/span& &span class=&mi&&512&/span& &span class=&o&&*&/span& &span class=&mi&&10&/span& &s}