pi-deeplearning / raspberrypi-facedetection-mtcnn-caffe-with-motion Goto Github PK

Hi,

@solderzzc @DuinoDu @VictorLULI
could you please provide train script and mention the dataset which you used to train P-R and ONet?

what is the different between RNet, PNet and ONet? I assume ONet alone gives the best results as it is deeper, doesn't it?

I got a problem that when import caffe on my RPI 3B, the usage of CPU is about 75% high.
I open a terminal, just type 'python', then type 'import caffe', and got CPU usage 75%....
I don't know what is wrong.
Have you ever meet this problem?

Hi,

when I run demo.py, I get the following error:
TypeError: 'numpy.float64' object cannot be interpreted as an index

Caffe has been compiled successfully and there was not problem in caffe compilation.

Hi,

I have tried an image with 720HD resolution and it took 5 seconds to detect faces. How one can speed up detection in this algorithm?

Hi,
@solderzzc @DuinoDu @VictorLULI
is there any way to get advantage of built-in GPU in raspberry pi in order to speed up the detection?

Did you only use the cpu of pi3 to run MTCNN ?

Hi,

as model is big for limited GPU memories, is it possible to compress the model by binarization or quantization method not only to reduce the size of models, but also to speed up the process?

pi@raspberrypi:~/RaspberryPi-FaceDetection-MTCNN-Caffe-With-Motion $ ./Demo.ipynb
./Demo.ipynb: line 2215: syntax error: unexpected end of file
Somebody can help me? thanks !