童年的梦想,一生的追求--兰迪最后的演讲2(转)

童年的梦想,一生的追求--兰迪最后的演讲()2

But it’s not just our bosses, we learn from our students.  I think the best head fake of all time comes from Caitlin Kelleher.  Excuse me, Doctor Caitlin Kelleher, who just finished up here and is starting at Washington University, and she looked at Alice when it was an easier way to learn to program, and she said, yeah, but why is that fun? I was like, ‘cause uh, I’m a compulsive male… like to make the little toy soldiers move around by my command, and that’s fun. She’s like, hmm.  And she was the one who said, no, we’ll just approach it all as a storytelling activity. And she’s done wonderful work showing that, particularly with middle school girls, if you present it as a storytelling activity, they’re perfectly willing to learn how to write computer software.  So all-time best head fake award goes to Caitlin Kelleher’s dissertation.

不但是领导可以给我知识,学生一样给我知识。我认为有史以来最善于使用障眼法的是凯特琳?凯莱赫。对不起,凯特琳?凯莱赫博士。她刚从本校毕业,开始在华盛顿大学工作。当时,爱丽丝软件只是一个提供学习编程的简易软件,她看着这个软件问:啊,这有什么好玩的?我觉得挺有意思,便说:因为,啊,我是个容易着迷的男生… …喜欢指挥玩具士兵走来走去,这就很有意思了。她也觉得挺有意思。嗯,他接着说,不,我们应该研发这个产品,让这个产品象是在给人讲故事一样。后来,他出色的完成了这个工作。同时表明,尤其是对女中学生来说,如果你把编程作为一个讲故事的活动,他们非常愿意学习如何写计算机软件。因此,毕业论文中,有史以来最佳障眼法奖将颁发给凯特琳?凯莱赫。

President Cohen, when I told him I was going to do this talk, he said, please tell them about having fun, because that’s what I remember you for. And I said, I can do that, but it’s kind of like a fish talking about the importance of water.  I mean I don’t know how to not have fun.  I’m dying and I’m having fun. And I’m going to keep having fun every day I have left.  Because there’s no other way to play it.

这是科恩校长。当我告诉他我要做这个讲座时,他说:请告诉他们尽享快乐,因为我记得你就是这样的。我说:我能做到,不过这有点像一条鱼讲述水的重要性一样。我的意思是我不知道怎么让自己不快乐。我会死去,但是我依然快乐地生活。我还会继续快乐地度过余生中的每一天,因为快乐是唯一的生活方式。

So my next piece of advice is, you just have to decide if you’re a Tigger or and Eeyore. [shows slide with an image of Tigger and Eeyore with the phrase “Decide if you’re Tigger or Eeyore”] I think I’m clear where I stand on the great Tigger/Eeyore debate. [laughter] Never lose the childlike wonder.  It’s just too important.  It’s what drives us.

因此,我的下一条建议就是, 你得决定你是做个快乐的跳跳虎呢,还是做一个悲哀的依哟驴。 [投影显示:跳跳虎和依哟驴的图片, 旁白是决定要做跳跳虎还是做依哟驴”] 我认为,我已经明确地表明了我对这场跳跳虎和依哟驴之大辩论的立场。[笑声] 绝对不要失去了自己的童心,它太重要了,它可以促使我们前进。

Help others.  Denny Proffitt knows more about helping other people.  He’s forgotten more than I’ll ever know.  He’s taught me by example how to run a group, how to care about people.

帮助别人。丹尼?普罗菲特更善于助人为乐。或许他已经忘记了,但是我记得非常清楚。他曾身体力行地教我如何带领团队,如何关心他人。

M.K. Haley – I have a theory that people who come from large families are better people because they’ve just had to learn to get along.  M.K. Haley comes from a family with 20 kids. [audience collectively “aaahs”] Yeah.  Unbelievable.

这是M?K?哈莉——我有一个理论,即:来自大家庭的人更好,因为他们必需学会与别人和睦相处。M?K?哈莉一家有二十个兄弟姐妹。 [听众发出"啧啧"] 恩,难以置信。

And she always says it’s kind of fun to do the impossible.  When I first got to Imagineering, she was one of the people who dressed me down, and she said, I understand you’ve joined the Aladdin Project. What can you do? And I said, well I’m a tenured professor of computer science. And she said, well that’s very nice Professor Boy, but that’s not what I asked.  I said what can you do? [laughter]

她总是说,挑战极限,其乐无穷。当我首次到迪士尼幻想工程的时候,她就曾经教育过我。她说:我知道你已经加入阿拉丁项目,你能做什么呢?我回答道:可我是个计算机科学终身教授啊。她则说:恩,不错嘛,教授先生,可我问的不是这个。我是问,你能做什么?” []

And you know I mentioned sort of my working class roots. We keep what is valuable to us, what we cherish. And I’ve kept my letterman’s jacket all these years.  I used to like wearing it in grad school, and one of my friends, Jessica Hodgins would say, why do you wear this letterman’s jacket? And I looked around at all the non­athletic guys around me who were much smarter than me. And I said, because I can. [laughter] And so she thought that was a real hoot so one year she made for me this little Raggedy Randy doll. [takes out Raggedy Randy] [laughter] He’s got a little letterman’s jacket too. That’s my all-time favorite.  It’s the perfect gift for the egomaniac in your life.  So, I’ve met so many wonderful people along the way.

各位,刚才我提到的是自己作为工薪阶层的某些本性。我们保存着对自己来说弥足珍贵的东西。这么多年来,我还一直保存着自己的优秀运动员外套。上研究生的时候,我最喜欢穿它。可我的一个朋友杰西卡?霍金斯就问我:你为什么要穿这件优秀运动员外套?我看了看周围那些不爱运动,但比我要聪明得多的人说:因为我能。”[笑声] 她认为这太搞笑了,于是有一年,她就给我做了这个小破兰迪玩具娃娃。 [拿出小破兰迪玩具娃娃来] [笑声] 你看,他也有一个小优秀运动员外套。这是我的最爱。对一个自大狂来说,这是一生中最好的礼物。所以,我在人生道路上遇到了很多的良师益友。

Loyalty is a two way street. There was a young man named Dennis Cosgrove at the University of Virginia, and when he was a young man, let’s just say things happened. And I found myself talking to a dean.  No, not that dean. And anyway, this dean really had it in for Dennis, and I could never figure out why because Dennis was a fine fellow. But for some reason this Dean really had it in for him.  And I ended up basically saying, no, I vouch for Dennis. And the guy says, you’re not even tenured yet and you’re telling me you’re going to vouch for this sophomore or junior or whatever?  I think he was a junior at the time.  I said, yeah, I’m going to vouch for him because I believe in him.  And the dean said, and I’m going to remember this when your tenure case comes up. And I said, deal.  I went back to talk to Dennis and I said, I would really appreciate you… that would be good.  But loyalty is a two-way street. That was god knows how many years ago, but that’s the same Dennis Cosgrove who’s carrying Alice forward.  He’s been with me all these years. And if we only had one person to send in a space probe to meet an alien species, I’m picking Dennis. [laughter] You can’t give a talk at Carnegie Mellon without acknowledging one very special person. And that would be Sharon Burks.  I joked with her, I said, well look, if you’re retiring, it’s just not worth living anymore.  Sharon is so wonderful it’s beyond description, and for all of us who have been helped by her, it’s just indescribable.  I love this picture because it puts here together with Syl, and Syl is great because Syl gave the best piece of advice pound-for-pound that I have ever heard.  And I think all young ladies should hear this. Sil said, it took me a long time but I’ve finally figured it out.  When it comes to men that are romantically interested in you, it’s really simple.  Just ignore everything they say and only pay attention to what they do. It’s that simple.  It’s that easy. And I thought back to my bachelor days and I said, damn. [laughter]

忠诚是相互的。在弗吉尼亚大学,有个年轻人叫丹尼斯?科斯格罗夫,那时候他还年轻,我只说事情吧。当时,我在跟院长谈话。不,不是现在那个院长啊。不管怎样,这院长真的跟丹尼斯有点纠葛。我一直都弄不懂为何,因为丹尼斯是个不错的人。但出于某种原因,这院长就想整他。谈到最后,我就直截了当地说:不能,我给丹尼斯担保。这个家伙就说:你连终身教职都不是,你竟然要告诉我,要给这个大二的学生担保?是大二还是大三来着?那时可能是大三吧。我说:是,我给他担保,我信任他。这院长就说:好,我记住了,等你评终身教职时我都记得。我说:成。我回去告诉丹尼斯说:我很真的很欣赏你……那件事情搞定了。但忠诚是相互的。这在多年前,天条一般的道理;而现在,带着同样的信念,丹尼斯?科斯格罗夫在推动着爱丽丝软件的发展。这么多年,他一直在我左右。如果我们用宇宙飞船只送一个人去与外星人会面的话,我选丹尼斯。 [笑声] 在此演讲,我必须对一个人表达我特殊的谢意,那就是莎伦?伯克丝。我曾跟她开玩笑说:唉,如果你退休的话,活着就没有意义了。莎伦非常优秀,简直难以言表;对我们这些受其恩泽的人来说,她的伟大,无法形容。我喜欢这张照片,因为西尔也在照片上。西尔的出色在于,她给了我货真价实的最好的建议。我想所有的女士都应该听听,西尔说:花了很长时间,我终于搞明白了:如果要跟男人谈恋爱,很简单;不管他说什么,只管他做什么。就那么简单。回想起我的单身生活,我说: [笑声]

Never give up.  I didn’t get into Brown University.  I was on the wait list.  I called them up and they eventually decided that it was getting really annoying to have me call everyday so they let me in. At Carnegie Mellon I didn’t get into graduate school. Andy had mentored me.  He said, go to graduate school, you’re going to Carnegie Mellon. All my good students go to Carnegie Mellon.  Yeah, you know what’s coming. And so he said, you’re going to go to Carnegie Mellon no problem. What he had kind of forgotten was that the difficulty of getting to the top Ph.D. program in the country had really gone up.  And he also didn’t know I was going to tank my GRE’s because he believed in me. Which, based on my board scores was a really stupid idea.  And so I didn’t get into Carnegie Mellon.  No one knows this.  ‘Til today I’m telling the story.  I was declined admission to Carnegie Mellon. And I was a bit of an obnoxious little kid. I went into Andy’s office and I dropped the rejection letter on his desk.  And I said, I just want you to know what your letter of recommendation goes for at Carnegie Mellon. [laughter] And before the letter had hit his desk, his hand was on the phone and he said, I will fix this. [laughter] And I said, no no no, I don’t want to do it that way.  That’s not the way I was raised. [In a sad voice] Maybe some other graduate schools will see fit to admit me. [laughter] And he said, look, Carnegie Mellon’s where you’re going to be. He said, I’ll tell you what, I’ll make you a deal.  Go visit the other schools.  Because I did get into all the other schools.  He said, go visit the other schools and if you really don’t feel comfortable at any of them, then will you let me call Nico?  Nico being Nico Habermann and I said, OK deal. I went to the other schools. Without naming them by name -- [in a coughing voice] Berkeley, Cornell.  They managed to be so unwelcoming that I found myself saying to Andy, you know, I’m going to get a job. And he said, no, you’re not. And he picked up the phone and he talked in Dutch. [laughter] And he hung up the phone and he said, Nico says if you’re serious, be in his office tomorrow morning at eight a.m.  And for those of you who know Nico, this is really scary.  So I’m in Nico Habermann’s office the next morning at eight a.m. and he’s talking with me, and frankly I don’t think he’s that keen on this meeting.  I don’t think he’s that keen at all. And he says, Randy, why are we here? And I said, because Andy phoned you? Heh. [laughter] And I said, well, since you admitted me, I have won a fellowship. The Office of Naval Research is a very prestigious fellowship.  I’ve won this fellowship and that wasn’t in my file when I applied. And Nico said, a fellowship, money, we have plenty of money.  That was back then. He said, we have plenty of money. Why do you think having a fellowship makes any difference to us?  And he looked at me.  There are moments that change your life. And ten years later if you know in retrospect it was one of those moments, you’re blessed. But to know it at the moment…. With Nico staring through your soul. [laughter] And I said, I didn’t mean to imply anything about the money.  It’s just that it was an honor.  There were only 15 given nationwide. And I did think it was an honor that would be something that would be meritorious.  And I apologize if that was presumptuous. And he smiled. And that was good.

永不言弃。我当时没有被布朗大学录取。我在候选名单上。我就每天给他们打电话,结果他们烦了,就把我录了。在卡内基?梅隆大学,我也没有被研究生院录取。安迪是我的导师,他说:去读到研究生,你要去卡内基?梅隆大学,我所有的好学生都去卡内基?梅隆大学。嗯,各位都知道下文如何。他说:你去卡内基?梅隆大学没问题。他或许不知道,获得国内顶尖博士学位难于登天。他也不知道,我的研究生入学考试考得一塌糊涂,因为他相信我。就我的考试分数而言,读博士的这个想法相当愚蠢。所以卡内基?梅隆大学没有录取我。没有人知道这一点。直的今天,我讲这个故事之前,没有人知道我曾被卡内基?梅隆大学拒绝过。我那时是个烦人的小孩,我走进安迪的办公室,把拒绝录取的信件扔在他桌子上,说:我只想让你知道你的推荐信在卡内基梅隆大学的份量。” [笑声] 信还没落到桌上,他就拿起电话说:我来修理他们。” [笑声] 我说:别,别,别,我可不想这样做,这样有失我的身份。[做悲哀状] 或许有其他研究生院觉得合适,并会录取我。” [笑声] 他说:不,卡内基?梅隆就是你该去的地方。他说:我告诉你怎么回事,跟你订个协议,去其他学校看看。因为我的确被其他的所有学校录取了。他接着说:去其他学校看看,如果你觉得在那些学校呆着不舒服,你再让我给尼科打个电话?尼科就是尼科?哈伯曼。我说:行,就怎么定了。我去了其他学校。就没有必要不说学校的名字了吧—— [做咳嗽声] 伯克利,康奈尔。这些学校让我觉得我不受欢迎,结果我跟安迪说:恩,我想找份工作。他说:不,你不用。他拿起话筒,用荷兰语打电话。 [] 他一挂电话就说:尼科说,认真点,明天上午八点去办公室找他。对你们这些人了解尼科的人,这实在恐怖。于是,第二天上午八点,我去了尼科?哈伯曼的办公室,他跟我谈话。坦白的讲,我觉得他并不想跟我会面,他一点也不热心。他问我:兰迪,来干嘛?我说:因为安迪给你打了电话?嘿嘿。” [] 我说:自从你接受我的申请后,我有赢得一份奖学金,海军研究办公室提供的一份不错的奖学金。我获得了这份奖学金,而我的申请材料上没有写。尼科说:奖学金,钱,我们有的是钱。这样就把我话挡了回去。他说:我们有足够的钱。为什么你觉得拿奖学金有那么重要呢?

来自 “ ITPUB博客 ” ,链接:http://blog.itpub.net/7201003/viewspace-524918/,如需转载,请注明出处,否则将追究法律责任。

转载于:http://blog.itpub.net/7201003/viewspace-524918/

  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
1 目标检测的定义 目标检测(Object Detection)的任务是找出图像中所有感兴趣的目标(物体),确定它们的类别和位置,是计算机视觉领域的核心问题之一。由于各类物体有不同的外观、形状和姿态,加上成像时光照、遮挡等因素的干扰,目标检测一直是计算机视觉领域最具有挑战性的问题。 目标检测任务可分为两个关键的子任务,目标定位和目标分类。首先检测图像中目标的位置(目标定位),然后给出每个目标的具体类别(目标分类)。输出结果是一个边界框(称为Bounding-box,一般形式为(x1,y1,x2,y2),表示框的左上角坐标和右下角坐标),一个置信度分数(Confidence Score),表示边界框中是否包含检测对象的概率和各个类别的概率(首先得到类别概率,经过Softmax可得到类别标签)。 1.1 Two stage方法 目前主流的基于深度学习的目标检测算法主要分为两类:Two stage和One stage。Two stage方法将目标检测过程分为两个阶段。第一个阶段是 Region Proposal 生成阶段,主要用于生成潜在的目标候选框(Bounding-box proposals)。这个阶段通常使用卷积神经网络(CNN)从输入图像中提取特征,然后通过一些技巧(如选择性搜索)来生成候选框。第二个阶段是分类和位置精修阶段,将第一个阶段生成的候选框输入到另一个 CNN 中进行分类,并根据分类结果对候选框的位置进行微调。Two stage 方法的优点是准确度较高,缺点是速度相对较慢。 常见Tow stage目标检测算法有:R-CNN系列、SPPNet等。 1.2 One stage方法 One stage方法直接利用模型提取特征值,并利用这些特征值进行目标的分类和定位,不需要生成Region Proposal。这种方法的优点是速度快,因为省略了Region Proposal生成的过程。One stage方法的缺点是准确度相对较低,因为它没有对潜在的目标进行预先筛选。 常见的One stage目标检测算法有:YOLO系列、SSD系列和RetinaNet等。 2 常见名词解释 2.1 NMS(Non-Maximum Suppression) 目标检测模型一般会给出目标的多个预测边界框,对成百上千的预测边界框都进行调整肯定是不可行的,需要对这些结果先进行一个大体的挑选。NMS称为非极大值抑制,作用是从众多预测边界框中挑选出最具代表性的结果,这样可以加快算法效率,其主要流程如下: 设定一个置信度分数阈值,将置信度分数小于阈值的直接过滤掉 将剩下框的置信度分数从大到小排序,选中值最大的框 遍历其余的框,如果和当前框的重叠面积(IOU)大于设定的阈值(一般为0.7),就将框删除(超过设定阈值,认为两个框的里面的物体属于同一个类别) 从未处理的框中继续选一个置信度分数最大的,重复上述过程,直至所有框处理完毕 2.2 IoU(Intersection over Union) 定义了两个边界框的重叠度,当预测边界框和真实边界框差异很小时,或重叠度很大时,表示模型产生的预测边界框很准确。边界框A、B的IOU计算公式为: 2.3 mAP(mean Average Precision) mAP即均值平均精度,是评估目标检测模型效果的最重要指标,这个值介于0到1之间,且越大越好。mAP是AP(Average Precision)的平均值,那么首先需要了解AP的概念。想要了解AP的概念,还要首先了解目标检测中Precision和Recall的概念。 首先我们设置置信度阈值(Confidence Threshold)和IoU阈值(一般设置为0.5,也会衡量0.75以及0.9的mAP值): 当一个预测边界框被认为是True Positive(TP)时,需要同时满足下面三个条件: Confidence Score > Confidence Threshold 预测类别匹配真实值(Ground truth)的类别 预测边界框的IoU大于设定的IoU阈值 不满足条件2或条件3,则认为是False Positive(FP)。当对应同一个真值有多个预测结果时,只有最高置信度分数的预测结果被认为是True Positive,其余被认为是False Positive。 Precision和Recall的概念如下图所示: Precision表示TP与预测边界框数量的比值 Recall表示TP与真实边界框数量的比值 改变不同的置信度阈值,可以获得多组Precision和Recall,Recall放X轴,Precision放Y轴,可以画出一个Precision-Recall曲线,简称P-R
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值