Roronoa-Zoro
Roronoa-Zoro
数据集格式，x1,y1,x2,y2,x3,y3,x4,y4，x,y,w,h 都是原始坐标，没有进行归一化
我知道错在那里了，数据类别写错了，数据集里面的数据我把它映射到0，1，2这种形式了，换成原始字符串就能训练了。
我也遇到了这个问题,类别修改过了.还有啥解决方法吗?
I also meet this problems, I solved it without "ImgSplit_multi_process" . My image shape is 1024*1024, so i don`t need to split img . So I just changed the file's...
You can try changing 'from preprocessing import ssd_vgg_preprocessing’ to 'import ssd_vgg_preprocessing', because they're in the same folder。In the same way，in ssd_vgg_preprocessing.py, turn 'from preprocessing import tf_image' into 'import tf_image'. Good...
Thanks for your reply. It helps a lot, but I am still a little confused. Let me explain my understanding first. In inference stage, colmap is used to calculate the...
OK, thanks. This really solved my confusion.