from transform import train_patch, predict_patch, train_filter, predict_filter
batch_size = 2200
nb_epoch = 40
patch_step = 1
nb_filters = 16
nb_conv = 3
patch_size = 64
patch_step = 1
spath = '/home/beams/YANGX/cnn_prj_enhance/tf_prd_battery_20170501/'
ipath = 'weights/tf_mouse.h5'
wpath = 'weights/tf_battery.h5'
proj_start = 1200
proj_end = 1201
ind_tomo = range(proj_start, proj_end)
fname = '/home/beams1/YANGX/cnn_prj_enhance/battery1_ds/prj_00000.tiff'
#
# imgx = dxchange.read_tiff('/home/beams1/YANGX/cnn_prj_enhance/battery1_train/trainx.tif')
# imgy = dxchange.read_tiff('/home/beams1/YANGX/cnn_prj_enhance/battery1_train/trainy.tif')
#
# mdl = train_patch(imgx, imgy, patch_size, 3, nb_filters, nb_conv, batch_size, nb_epoch, ipath)
# mdl.save_weights(wpath)
img_n = dxchange.read_tiff_stack(fname, ind_tomo, digit = 5)
predict_patch(img_n, patch_size, 1, nb_filters, nb_conv, batch_size, wpath, spath)
# mdl.save_weights(wpath) # predict_patch(imgx3, patch_size, 1, nb_filters, nb_conv, batch_size, wpath, spath) print imgx.shape, imgy.shape mdl = train_patch(imgx, imgy, patch_size, 2, nb_filters, nb_conv, batch_size, nb_epoch) wpath = 'weights/400ms_new.h5' mdl.save_weights(wpath) fname = '/home/beams1/YANGX/cnn_prj_enhance/exposure_tests/C3S_insitu_3_181prj_400ms_248cycles_000.h5' data = dxchange.read_hdf5(fname,'/exchange/data') dark = dxchange.read_hdf5(fname,'/exchange/data_dark') white = dxchange.read_hdf5(fname, '/exchange/data_white') print data.shape, dark.shape, white.shape print data.min(), data.max() data1 = tomopy.normalize(data, white, dark) print data1.min(), data1.max() print data1.shape data1[data1>1] = 0.78 data1[data1<0] = 0 data1 = ds(data1) predict_patch(data1, patch_size, 1, nb_filters, nb_conv, batch_size, wpath, spath) # mdl = train_patch(imgx5[0], imgy, patch_size, 2, nb_filters, nb_conv, batch_size, nb_epoch, ipath) # wpath = 'weights/500ms.h5' # mdl.save_weights(wpath) # predict_patch(imgx5, patch_size, 1, nb_filters, nb_conv, batch_size, wpath, spath)