1709.08325
Reid问题指的是判断一个probe person 是否在被不同的camera捕获的gallery person 中出现。
通常是如下情景:给出一个特定camera下某个特定人的probe image 或者 video sequence,从其他camera处询问这个人的图像,地点,时间戳。
Key:
(1). Pose-driven, body part alignment, combine whole feature and body part feature, focus on alignment of part model,
(2). Combine image label and human attributes classes, do classification with attributes and identity learning
(3). Based on triplet loss, improve metric learning for an end to end learning
(4). Post-process, re-ranking
AlignedReID: Surpassing Human-Level Performance in Person Re-Identification
Hydraplus-net: Attentive deep features for pedestrian analysis.
Darkrank: Accelerating deep metric learning via cross sample similarities transfer.
Glad: Global-local-alignment descriptor for pedestrian retrieval.
由于发现cuda c++ 的 debug方式和c++ 差别很大,因此打算再开一篇,专门记录一些和error checking 以及debug有关的内容.
Error checks in CUDA code can help catch CUDA errors at their source. There are 2 sources of errors in CUDA source code:
update:有毒吧,kernel 中出问题原来不会报错。
请教了组里的 hust 学长 orz..
学到了 cuda-memcheck 命令和 cudaGetLastError 来查看问题,可以参考What is the canonical way to check for errors using the CUDA runtime API?
先放一波资料。
An Even Easier Introduction to CUDA CUDA C/C++ Basics nvidia-thrust 官方文档 how-access-global-memory-efficiently-cuda-c-kernels efficient-matrix-transpose-cuda-cc 很强大的warp内shuffle cuda-GDB官方文档 cuda-c-best-practices-guide cuda 提出的目的是能够让程序员透明地使用GPU来高效地进行并行运算。