HappyKerry

Ubuntu 16.04 cuda8.0 cudnn5.1测试ssd mobilenet，GPU模式，DLOG统计出的内存占用24.59M，GPU占用显存41.14M，系统监视器显示占用内存439.5M，nvidia-smi显示占用显存291M，为什么差这么多？

Why after the iteration of 20000 times, the loss oscillates around 1.4, no longer reduces

1

I used 512 ssd training, iterating about 20,000 times, the loss dropped to 1.4, the map is about 0.9, but then the loss starts to oscillate between 1-2.5, and the...

detectron net create error

5

I changed detectron pkl model to caffe2 pb model, errors happened when called CAFFE_ENFORCE(workspace.CreateNet(model.predict.net)); in imagenet.cc terminate called after throwing an instance of 'caffe2::EnforceNotMet' what(): [enforce fail at operator.cc:185] op....

support face alignment before recog?

4

After get keypoints , do face alignment before recog ,how to? thanks

yolov3 模型加速后目标丢失、可信度降低

10

简单对比了几张图，发现darket的检测结果与加速后模型检测结果不一致，tensorRT的检测会丢目标，甚至会丢darknet下可信度为1的目标，可能是什么原因

I want to detect object like license plate by using parallelogram， how to detect the four keypoints？