2016 年是我第一次参加 CVPR. 我以 FAIR 实习生的身份注册, 在会场和旷视的伙伴们 Tim, 舒畅汇合, 一起搭起了旷视的展台. 那一年, 旷视是 CVPR 的最高等级赞助商, 展台上展览了我参与的量化 CNN 工作 DoReFa-Net, Scott Gray 还来交流了里面的量化 kernel 要怎么写. DoReFa-Net 是旷视第一次在 FPGA 芯片上跑起 CNN, 这些技术积累成了十年后上市的爱芯元智的雏形. 我们隔壁, 是侯晓迪带着同事们在布置图森的展台.
在那一年的拉斯维加斯, 华人圈子里讨论最多的 gossip 是:"听说孙剑要去旷视了". 孙剑中途来旷视的展台跟我们打了个招呼, 懵懂的我当时还并不了解这个名字的含金量. 那一年, ResNet 毫不意外的获得了最佳论文奖. 在 CVPR 后的一个月, 恺明加入了 FAIR, 在公司安排的公寓里和我做了一段时间邻居. 再往后的几年中, 以恺明、Ross、Piotr 等人为核心的这个叫做 "win vision" 的团队, 是 CV 界的 best team in the world.
How To Do Ablation Experiments
延续 上一篇文章, 再说一说怎么科学的在 paper 里做 ablations.
Where Are Pixels? -- a Deep Learning Perspective
Technically, an image is a function that maps a continuous domain, e.g.
a box array[H][W], where each element
array[i][j] is a pixel.
How does discretization work? How does a discrete pixel relate to the abstract notion of the underlying continuous image? These basic questions play an important role in computer graphics & computer vision algorithms.
This article discusses these low-level details, and how they affect our CNN models and deep learning libraries. If you ever wonder which resize function to use or whether you should add/subtract 0.5 or 1 to some pixel coordinates, you may find answers here. Interestingly, these details have contributed to many accuracy improvements in Detectron and Detectron2.
Deep Learning Experiments and Claims
这几年来, 从 FAIR 的几位大佬身边学习到的最多的是对待 research 的态度. 因此说说写 paper 和做实验的体会.
实验是为了证明或强化文章里给出的 claim/hypothesis 的.
Ross ICCV 2019 tutorial 最后谈了谈怎么写 paper. 第 126 页说, 文章中所有的 claim, 理想情况下都应该 要么是参考文献中已有的, 要么是被实验证明的.
Fight Against Silent Bugs in Deep Learning Libraries
TL;DR: How to find out if your favorite deep learning library is occasionally giving you wrong results? Such bugs happen from time to time, and are extremely difficult to notice, report, and debug.
2018 Geekpwn Adversarial Attacks on Face Recognition
安全比赛 Geekpwn 今年开始强调 "人工智能安全", 办了一个 CAAD 对抗样本攻防赛. 这让我在多年没碰 CTF 之后又有了一个 CTF 的机会, 只是这次攻击的是 ImageNet 分类器..
Unawareness of Deep Learning Mistakes
TL;DR: People are hardly aware of any deep learning mistakes they made, because things always appear to work, and there are no expectations on how well they should work. The solution is to try to accurately reproduce settings & performance of high-quality papers & code.
这个领域里, 什么都特别快.
三个月前看到 Bengio 组的 BinaryConnect. 脸草的同事都很喜欢模型加速 / 压缩的主题, 因此立刻就重现了结果开始改进. 当时就说要做成 Binary Activation, 并且搞一个 GPU runtime. 正当同事们回家过年, 我在 yy 这学期 parallel 大作业要不就写这个 runtime 的时候, 昨天看到 Bengio 新的 paper 挂出来, 已经都做完了. 更夸张的是, 在前天 arxiv 挂了另一篇文章, 方法基本一样.
三个月, 能专心做的话并不难, 然而我要应付作业, 要去 oculus 写代码, 还有其他好玩的东西在分心. 想着有空慢慢做的时候, 别人已经不等你了.
OpenPano: How to write a Panorama Stitcher
This is a summary of the algorithms I used to write OpenPano: an open source panorama stitcher. You can find the source code on github.