search

anson0910 / CNN_face_detection

Implementation based on the paper Li et al., “A Convolutional Neural Network Cascade for Face Detection, ” 2015 CVPR
253 stars 148 forks source link

About the result after running #25

Closed MaxLingwei closed 8 years ago

MaxLingwei commented 8 years ago

Hello, I downloaded you project and ran in my computer. Your work helps me a lot. A problem is that my result is shown as follow: 2

Of course it's not good enough. I wonder it is because of my incorrect operation, or there are some bugs in your code/model? What is the result you run in you own environment? I never change your code or your caffee model. Hope you would answer my question. Thanks a lot.

anson0910 commented 8 years ago

Hi, you can refer to this issue. The current parameters (thresholds) are not tuned at their best values, so I guess you can try tuning the parameters, for example : increasing the confidence value in the detect_face_48c method of face_detection_functions.py

Some other possibilities that cause these issues is there was an error during training of the calibration nets, as mentioned here. The data preparation scripts have been changed as well as mentioned here.

MaxLingwei commented 8 years ago

Thank you very much. By the way, what dataset do you use to train the calibration nets? Do you use all face in aflw? And how about the number of positive and negative train sets?

anson0910 commented 8 years ago

Yes, I use all faces in AFLW, total images of 45 categories adding up to roughly 1 million training images.

MaxLingwei commented 8 years ago

I get a better result after I have modify the confidence and threshold. Later I will train a caffeemodel by myself. Thank you for your kindness!