StephanPan

FacialComponentDiscriminator

1

FacialComponentDiscriminator是一个全卷积网络，输出是一个20*20的张量，似乎和常规的判别器形式不太一致，为什么没有用全连接层使结果以标量的形式输出呢？

where is the efficientnet pretrained model [Usage]

2

**Please use this template to submit your problem or your question will _NOT_ be properly tackled.** --- Describe what you want to do, including: i want to use efficientnet as...

为什么用wh ratio做target assignment？？

1

build_targets中用wh ratio做target assignment吗？感觉应该用iou做啊。

To transfer the motion blur between the image pair (x, y) to a new image xˆ, we can simply compute: yˆ := F(ˆx, G(x, y)). however, when i set the...

some module missing in benchmark.py

3

No module named 'dl_lib.config'

is it correct to get video face swapping by inferring video frame by frame?

question for uniform matcher

9

according to your code, the uniform matcher seems calculate the L1 distance between pred_bbox/anchor with target among batch imgs. but i think it should be computed within single img. another...

Autoencoder loss of step1

6

I use the labelenc in a custom model. In training step1 the autoencoder loss are not going down。in my opinion，the autoencoder loss should be in sync with detection loss since...

gpu memory and latency time

3

During the model warm-up, especially in the initial steps, there is significant fluctuation in GPU memory, and the process takes a considerable amount of time, reaching around twenty seconds. The...

when i use real expression coefficients extracted by deep3d, the performance drops a lot. why?