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 147 forks source link

About the training step #27

Closed MaxLingwei closed 8 years ago

MaxLingwei commented 8 years ago

Dear anson: I have ran you code before and now I'd like to train my own caffee model. But I don't understand the training step. For example, I don't know what training set I should use. According to my understanding, I should use 12*12 pixels image when I training the 12net. But in you code create_negative.py I find that you crop image to different size. I don't understand why you do so. So I do hope you can answer my question if convenient. Thank you very much!

anson0910 commented 8 years ago

Hi, For the training set, you can use the AFLW dataset for positive samples, and any images without faces for negative samples. The reason why different sizes of crops are taken from images for training is to create as many negative samples as possible from a single image, and caffe will automatically convert them to 12*12 pixel images if set properly in the prototxt file.

MaxLingwei commented 8 years ago

Oh, I see. Thank you very much. And I have another question. Why do you save negative samples into different directory? Why don't just save all negative samples in the same folder?

anson0910 commented 8 years ago

Oh, that's just for a silly reason, Nautilus on Ubuntu crashes when opening folders containing too many files, haha

MaxLingwei commented 8 years ago

Oh, I understand. Thank you very much. You help me a lot.