-
In the preprocessing, `CROP_PADDING` is used to pad the `image_size` to determine a ratio for center crop. Is there a particular reason for the choice of `CROP_PADDING`=32?
-
Dear @1e100 ,
Nice work!
According to my experience, training mnasnet will takes about 7~8G GPU memory for batch_size=96(mobilenetV2 setting), since i tried to reproduce paper values with pytorch …
-
## 🐛 Bug
## To Reproduce
Steps to reproduce the behavior:
In the demo dir, tried running:
"
python webcam.py --min-image-size 300 MODEL.DEVICE cpu
"
the following error was returned:
…
-
- vulkan version: [Vulkan_loader_demos_1.1.100.tar.gz](https://developer.nvidia.com/embedded/vulkan)
- glslang version
```sh
(nano) real2tech@real2tech-desktop:~/w/glslang$ glslangValidator --ve…
-
Hello,I have a resnet50 caffe model,and use ncnn do inference on armeabi-v7a, and my time is 1534ms. I use Caffe-Int8-Convert-Tools canvert my model to int8 and run inference,my time is 2631ms. It's s…
-
polaris:/data/local/tmp $ ./benchncnn 4 8 0 -1 …
-
- android-armv8-vulkan version
- code version: de707145, Wed Apr 10
logs as below:
```shell
chiron:/data/local/tmp/ncnn $ ./benchncnn 8 1 1 0
[0 Adreno (TM) 540] queueC=0 qu…
-
Hi, thank you for your work and for open sourcing your code!
As you don't list mail addresses in the paper I abuse the Github issue tracker to ask a question about the paper:
Have you compared yo…
-
What about MNasnet https://github.com/AnjieZheng/MNasnet-Pytorch?
bhack updated
5 years ago
-
mobilenet和mobilenetssd在树莓派3b上运行时间较reamde中的时间分别多了30ms和80ms,但是其他网络的运行时间基本正常。编译步骤和运行方式完全和指南中的一致,请问可能是什么原因造成mobilenet的运行时间过长?
下面是在树莓派3b上运行benchmark后的结果,
pi@raspberrypi:~/code/ncnn2/ncnn/build/benchm…