DeepFaceDrawing-Jittor icon indicating copy to clipboard operation
DeepFaceDrawing-Jittor copied to clipboard

How is your web version background deployed?

Open TomatoBoy90 opened this issue 4 years ago • 4 comments

I use flask for deployment, and the infer time is very slow, one second at a time

TomatoBoy90 avatar Sep 01 '20 01:09 TomatoBoy90

Most of the time is spent in inference.

TomatoBoy90 avatar Sep 01 '20 01:09 TomatoBoy90

I use a single GPU (P 100) for inference

TomatoBoy90 avatar Sep 01 '20 02:09 TomatoBoy90

Most of the time is spent on shadow synthesis. In our online system, the shadow is composed of sketch images. To speed up the reading and writing time, we save the sketch as a binary file. In this version, each sketch image needs to be reconstructed by the corresponding decoder. So more time is needed. Saving these images in advance can save a lot of time. Hope it can help you.

onlineplay avatar Sep 08 '20 13:09 onlineplay

Most of the time is spent on shadow synthesis. In our online system, the shadow is composed of sketch images. To speed up the reading and writing time, we save the sketch as a binary file. In this version, each sketch image needs to be reconstructed by the corresponding decoder. So more time is needed. Saving these images in advance can save a lot of time. Hope it can help you. thank you very much,I'm going to experiment for you suggest

TomatoBoy90 avatar Sep 09 '20 01:09 TomatoBoy90