Git Product home page Git Product logo

Comments (12)

cristianoc20 avatar cristianoc20 commented on July 30, 2024

因为是你没有改动我的代码的缘故 最后代码demo部分是soft-nms 效果很差 你把它改过来nms的方式应该会好很多 另外你测试的时候阈值也多测试一下 这个权重文件线上可以至少达到0.97左右

from rebar_detection.

JiangZongKang avatar JiangZongKang commented on July 30, 2024

好的,确实是这样,改过来就正常了

from rebar_detection.

cristianoc20 avatar cristianoc20 commented on July 30, 2024

恩恩

from rebar_detection.

JiangZongKang avatar JiangZongKang commented on July 30, 2024

还有一个问题,一般训练多少个epoch比较合适呢,我看yunyang代码是50代,你是300代,我按照300代训练,30代之后loss都是nan,对于这点我有点迷惑,还望解答我的困惑。

from rebar_detection.

cristianoc20 avatar cristianoc20 commented on July 30, 2024

这个要看效果 我一般是先训练一些小epoch 看看loss能不能降 然后用生成的权重文件测试一下看看效果 后面的话我就会直接训练长epoch 看看更大的epoch loss是不是能降低 如果已经用很大的epoch loss还是差不多那就是数据或者别的什么问题 已经降不下去了 nan的原因有挺多的 在这份代码里面最主要的有两个原因 第一个是我们是多尺度训练 很容易就nan 你可以把尺寸的范围减少几个 甚至先用单一尺度训练 看看是不是多尺度的问题 第二个是我们的数据可能经过augument之后 有些gt box的w h变成0 导致我们的除数变成了0 你在那几处加个tf.maximum(union_area,1e-12)就可以 当然原repo也说了一般会出现几代nan 后面就会变回正常 这个我自己也是有出现 过多几个epoch就正常了

from rebar_detection.

JiangZongKang avatar JiangZongKang commented on July 30, 2024

好的,我去试试,谢谢你的解答。

from rebar_detection.

Space-Xun avatar Space-Xun commented on July 30, 2024

你好,

我使用你分享的权重在200张测试集上预测结果并提交比赛网站后,得分是0.27左右,我想了一下午也没有找到哪里出问题了,所以向你请教下这是什么原因?
请问你是如何生成比赛所要求提交的csv文件的?

from rebar_detection.

cristianoc20 avatar cristianoc20 commented on July 30, 2024

0.27的原因在前面说了 把默认的soft-nms注释掉 用普通的nms即可

from rebar_detection.

Space-Xun avatar Space-Xun commented on July 30, 2024

0.27的原因在前面说了 把默认的soft-nms注释掉 用普通的nms即可
哦哦,那请问如何生成提交的csv文件?

from rebar_detection.

Space-Xun avatar Space-Xun commented on July 30, 2024

我看了一下没找到生成提交的csv文件的代码?

from rebar_detection.

cristianoc20 avatar cristianoc20 commented on July 30, 2024

提交代码在tool/commit.py,readme没说清楚抱歉哈

from rebar_detection.

Space-Xun avatar Space-Xun commented on July 30, 2024

谢谢!太客气啦

from rebar_detection.

Related Issues (5)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.