Open Dantju opened 4 years ago
@tensorboy 请问为什么evalute中的载入的val数量只有2693,coco val中总共有5000才对
明天回答你可以吗,因为只有2693张里面有人。
On Sun, Mar 8, 2020 at 10:45 PM Dantju notifications@github.com wrote:
@tensorboy https://github.com/tensorboy 请问为什么evalute中的载入的val数量只有2693,coco val中总共有5000才对
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/tensorboy/centerpose/issues/27?email_source=notifications&email_token=AEDERP2W7ARXZ36DHIP5IDDRGR7BHA5CNFSM4LD7PPE2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEOFXJUI#issuecomment-596341969, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEDERPYI36WOMVJR33HBMNDRGR7BHANCNFSM4LD7PPEQ .
@tensorboy ok 对了你的centerpose相对于centernet都做了哪些改进了 除了换了几个backbone之外
@Dantju 你好,我其实并没有太多的改进,就是调了调参数,DLA34对应的参数都在这里:https://github.com/tensorboy/centerpose/blob/master/experiments/dla_34_512x512.yaml
@tensorboy 调了参数之后AP有上升吗
当然有了,只要不断调就会越来越高。
On Mon, Mar 9, 2020 at 10:42 PM Dantju notifications@github.com wrote:
@tensorboy https://github.com/tensorboy 调了参数之后AP有上升吗
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/tensorboy/centerpose/issues/27?email_source=notifications&email_token=AEDERPYVCYR32FARLKSD66LRGXHK5A5CNFSM4LD7PPE2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEOKC5HY#issuecomment-596913823, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEDERP2E5Q44OF7URGOXR7LRGXHK5ANCNFSM4LD7PPEQ .
@tensorboy 主要调了哪些参数呢
如果都枚举出来的话,其实我也忘的差不多了。。。反正能看到不少提高的地方。。。
@tensorboy 请问一下centernet是有两个分支,一个PD,一个PE,你有没有发现同时训练PD_PE中的PD的效果是差于单独训练的PD的效果,你有研究如何改善这个问题吗,如果能改善PD_PE中的PD是否会对PE也有促进效果呢
PE和PD是什么的缩写呢。@Dantju
person detection 和 Pose Estimation
@tensorboy
你的centerpose不就是做PD和PE 你有用centetnet单独训练过PD吗
@tensorboy 请问主要调了哪些参数呢,哪些是主要的参数
@sulthanallaudeen this is scam. don't interact
@sulthanallaudeen this is scam. don't interact
Block and report this scammer account.
@tensorboy 你好,看你的表格里面DLA-34在coco val上的精度是62.7%,我这边用官方的model测出来是59.2%,请问你做了什么别的优化措施嘛