lvpchen

Results 7 issues of lvpchen

用自己的数据集训练出的模型,检测结果有时会出现置信度和类别为-1的情况,大部分情况下正常。 这是out输出结果:{'detection_out': array([[[[ 0., -1., -1., -1., -1., -1., -1.]]]], dtype=float32)} box:[[-1280 -720 -1280 -720]](我原图是1280x720) cls:[-1.] conf:[-1.] 请问下是什么原因导致的?

用自己训练出的模型测试,检测到目标后将目标快速移开,框还是在原来的位置。请问下这是什么原因。谢谢。。。

Hello, now I have some false detections images. I want to take its as negative samples, how to add its to train sets to retrain? Thank you.

多个gpu训练,一直卡在下面,请问下怎么回事? I1021 16:08:19.171383 4102 solver.cpp:70] Solver scaffolding done. I1021 16:08:19.176095 4102 caffe.cpp:255] Starting Optimization I1021 16:08:19.751611 4243 upgrade_proto.cpp:79] Attempting to upgrade batch norm layers using deprecated params: mobilenet_yolov3_lite_train.prototxt I1021 16:08:19.751641...

你好,请教你在转onnx模型时,将solov2_head.py中get_seg() 如下操作放入onnx model,这个对应要改哪个地方?

Hello, I do not know how to test? Can you upload your demo code. thanks

这个分支和yolov4那个分支有啥区别,也可以用来训练v4,v4-tiny吧