bingykang / Fewshot_Detection

Few-shot Object Detection via Feature Reweighting
https://arxiv.org/abs/1812.01866
528 stars 111 forks source link

模型训练完成后,推理时间只有3ms,但是box filter却有上百毫秒,请问是为什么呢。 #92

Open mazhenxuan opened 1 year ago

STSTERANDMOMO commented 1 year ago

你有跑出结果吗,我测试结果map都是0,我发现我的results文件夹下的comp4_det_test_的txt文件都是空的,你知道是什么问题吗

mazhenxuan commented 1 year ago

comp4_det_test_存放的是各类框的准确率以及坐标信息,如果是空的,应该是你的forward推理有问题。

马振轩 @.***

 

------------------ 原始邮件 ------------------ 发件人: "bingykang/Fewshot_Detection" @.>; 发送时间: 2023年4月9日(星期天) 下午2:43 @.>; @.**@.>; 主题: Re: [bingykang/Fewshot_Detection] 模型训练完成后,推理时间只有3ms,但是box filter却有上百毫秒,请问是为什么呢。 (Issue #92)

你有跑出结果吗,我测试结果map都是0,我发现我的results文件夹下的comp4_det_test_的txt文件都是空的,你知道是什么问题吗

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

STSTERANDMOMO commented 1 year ago

comp4_det_test_存放的是各类框的准确率以及坐标信息,如果是空的,应该是你的forward推理有问题。 马振轩 @.   ------------------ 原始邮件 ------------------ 发件人: "bingykang/Fewshot_Detection" @.>; 发送时间: 2023年4月9日(星期天) 下午2:43 @.>; @*.**@*.>; 主题: Re: [bingykang/Fewshot_Detection] 模型训练完成后,推理时间只有3ms,但是box filter却有上百毫秒,请问是为什么呢。 (Issue #92) 你有跑出结果吗,我测试结果map都是0,我发现我的results文件夹下的comp4_det_test_的txt文件都是空的,你知道是什么问题吗 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.>

可是我用的就是作者的代码啊,你做了什么改动么?

STSTERANDMOMO commented 1 year ago

你有没有你训练好的权重文件发我一个,我训练出来的是000180.weights,是不是我权重文件的问题?

QSJZDDY commented 1 year ago

你有没有你训练好的权重文件发我一个,我训练出来的是000180.weights,是不是我权重文件的问题?

你好能问一下你的环境配置吗,我训练的时候会报错 inconsistent tensor size, expected tensor [64 x 32 x 3 x 3] and src [1776] to have the same number of elements, but got 18432 and 1776 elements respectively at /pytorch/torch/lib/TH/generic/THTensorCopy.c:86 不知道你有没有遇到过这种问题