rbgirshick/rcnn

pre-computed models

hayderm opened this issue · 0 comments

Hi Ross,
Please, could you tell me how do you computed the binary models such as (finetune_ilsvrc13_val1+train1k_iter_50000,finetune_voc_2007_trainval_iter_70k,...) ? did you use caffe or Alex code ?
because I am using Alex code to generate the binary model so when I use rcnn_model.cnn.layers = caffe('get_weights'); after initialization with binary model computed by Alex code but it does not give the computed final weights learned from the trained model.