^o^
@eragonruan i use k40 Demo for /home/zhoulinyuan/project/text-detection-ctpn/data/demo/001.jpg Detection took 1.486s for 7 object proposals ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Demo for /home/zhoulinyuan/project/text-detection-ctpn/data/demo/002.jpg Detection took 1.524s for 7 object proposals ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Demo for /home/zhoulinyuan/project/text-detection-ctpn/data/demo/003.jpg Detection...
@whai362 Random sample may be cut off text lines, does this have no effect?
@whai362 I see, then when calculating labels, you calculate them on the original image, or you calculate the intersection point with the edge, then calculate the labels?
i use fpn as your done, but the result is bad. as far as you know, how much can FPN improve?
all we need is a segmentation result，kernel and the algorithm of expanding. ---------------------------------------------------------------------------------------------------------- your paper said: The backbone of PSENet is implemented from FPN, F = C(P 2 , P...
https://github.com/kpzhang93/MTCNN_face_detection_alignment/tree/master/code/codes/MTCNNv2 这个里面有的呀
应该在train-mtcnn发问题，发错地方了，它输入是24*24，这个还很重量级么，没有理解你的意思，能否测试一下精度，比如lfpw，test中有249张图片，能够告知mail么
我的邮箱：[email protected]，可以交流下么
> 那个模型有3M多 这个应该不算大吧，相对你其他的版本
模型名称 | 输入尺寸 | 计算量 | 训练时L2 | 训练时L1 | 备注 -- | -- | -- | -- | -- | -- Lnet_v1 | 48x48 | 3.8 M | 约0.0021...