bubbliiiing centernet-pytorch issues

bubbliiiing / centernet-pytorch

这是一个centernet-pytorch的源码，可以用于训练自己的模型。

MIT License

366 stars 98 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

按视频操作CUDA和pytorch不匹配

#63 kimtaesa opened 6 months ago
0
导儿，训练时候开启map值计算之后，下一轮训练卡住，已经定位是get map_txt函数中的问题导致，程序卡在数据加载的地方

#62 lkhwhkzz opened 10 months ago
4
total_c_loss=nan, total_r_loss=nan

#61 qq11-maker opened 10 months ago
0
导，我在训练时的验证集的mAP值，与重新单独测试的mAP值差距非常大，大概10个点左右，你遇到过佬？

#60 97-coffee opened 10 months ago
0
单张图片包含3000个单类别物体，检测发现map指标很低，是因为centernet算法检测数量限制吗？

#59 li-an666 opened 11 months ago
2
为什么训练到第六轮停止训练了

#58 lkhwhkzz opened 1 year ago
6
视频检测没有框

#57 tangpanyu opened 1 year ago
1
改动centernet的head

#56 YYL-SCAR opened 1 year ago
2
mAP计算过程中预测结果异常

#55 wkgit88 opened 1 year ago
10
mAP

#54 wkgit88 closed 1 year ago
0
关于hourglassnet

#53 Vision-LYF opened 1 year ago
0
修改input_shape后，还能使用centernet_resnet50_voc.pth预训练权重吗

#52 hua-zi closed 1 year ago
4
为什么用训练时用测试集验证map为67%，在get_map中用测试集验证达到81%，同一个数据集怎么差别这么大

#51 bukeN1 opened 1 year ago
7
怎么修改做下游任务比如关键点识别

#50 JoyWang1123 opened 1 year ago
1
关于summery.py中测试CenterNet_HourglassNet

#49 Vision-LYF opened 1 year ago
2
导儿，可以加入mosaic, mixup数据增强吗？

#48 Vision-LYF opened 1 year ago
2
请问如何输出mAP_75和mAP_90

#47 PhenixFeather opened 1 year ago
2
ResNet18权重文件

#46 wjl0512 opened 1 year ago
1
大佬您好！分类效果很差，都是一个类别是什么原因？

#45 ddddwx opened 1 year ago
8
请问我改mobilenetv3的时候运行到第7批次就自动停止了是怎么回事呢

#44 sunsn1997 opened 1 year ago
3
结果mAP很小很小，也就1%，这是什么原因啊

#43 ayangge opened 2 years ago
3
第一次尝试的新手提问

#42 Xie-Muxi-BK opened 2 years ago
1
二分类，训练的val数据的结果一直都是F1=0.17 ; Recall=9.09% ; Precision=100.00%？

#41 changgongcheng opened 2 years ago
6
RuntimeError: DataLoader worker (pid 57154) is killed by signal: Segmentation fault.

#40 yusirhhh opened 2 years ago
6
map指标

#39 ChristmasLee closed 2 years ago
2
评测指标

#38 winneryoow1 opened 2 years ago
1
Loss curve

#37 Kevin43614 opened 2 years ago
0
procession一直100%

#36 PeterSmith1 opened 2 years ago
1
显示no mudule named 'past'

#35 buloseshi closed 2 years ago
1
可以用来训练coco格式数据集吗？

#34 buloseshi closed 2 years ago
1
我将输入改为416X416，ap结果有个别类很差，map也很差

#33 tty0013 opened 2 years ago
7
关于和官方模型的区别

#32 CX-XC opened 2 years ago
1
冻结训练的问题

#31 Liqq1 opened 2 years ago
3
centernet本来用的是可变形卷积，你这个里面是不是用的简单的反卷积呀

#30 Lexie971102 opened 2 years ago
1
hourglassnet关于沙漏网络

#29 babyta opened 2 years ago
2
断点续传后loss值会上升很多，大概会白训练20个epoch

#28 seki18 opened 2 years ago
1
求教def gaussian2D中为什么要*h.max()？

#27 leesirgo opened 2 years ago
1
训练没有归一化，预测却有归一化，是不是有问题？

#26 seven-linglx closed 2 years ago
2
请问如何输出热力图？

#25 TCLHJ opened 2 years ago
1
你好

#24 wujinxiao opened 2 years ago
3
gaussian_radius里计算半径的公式，是否应该是除以2a

#23 seven-linglx opened 2 years ago
2
有没有办法输出每一轮的mAP？

#22 TCLHJ opened 2 years ago
3
请问大神有没有试过dsnt

#21 pengyang1225 opened 2 years ago
1
用代码给出的权重进行测试，结果mAP。。

#20 LeeSSangY3y opened 2 years ago
3
关于anchor-base改成anchor-free

#19 TCLHJ closed 3 years ago
1
训练自己的数据集AP22%

#18 TCLHJ closed 3 years ago
0
get_map运行后仍然是过去默认VOC20类的类别

#17 dodeokwangyooo opened 3 years ago
2
用coco工具计算0.50:0.95map提示Results do not correspond to current coco set

#16 zhiyuan454 opened 3 years ago
2
请问里面的yolo.py文件在哪？

#15 mlantern closed 3 years ago
2
改动残差块之后的疑惑

#14 chenpeng68 opened 3 years ago
2