#
#mnist-bg-fix_l2_20190125_220558_unet_adam_lr1e-05_wd0.0_batch12
#bn = 'mnist-fg-fix_l2_20190125_215745_unet_adam_lr1e-05_wd0.0_batch15'
#bn = 'mnist-fg-fix_l1smooth_20190127_000440_unet_adam_lr1e-05_wd0.0_batch12'
#bn = 'mnist-fg-fix_l1_20190126_093344_unet_adam_lr1e-05_wd0.0_batch12'
#bn = 'mnist-fg-fix_l1_20190126_093344_unet_adam_lr1e-05_wd0.0_batch12'
#bn = 'mnist-bg-fix_l2_20190125_220558_unet_adam_lr1e-05_wd0.0_batch12'
bn = 'mnist-fg-fix-v1_l2_20190128_165524_unet_adam_lr1e-05_wd0.0_batch48'
#bn = 'mnist-fg-fix-v2_l2_20190128_163917_unet_adam_lr1e-05_wd0.0_batch48'
model_path = log_dir_root_dflt / bn / 'checkpoint.pth.tar'
model = UNet(n_channels=n_ch, n_classes=n_ch)
state = torch.load(model_path, map_location='cpu')
model.load_state_dict(state['state_dict'])
model.eval()
#%%
# argkws = dict(output_size = 256,
# epoch_size = 10,
# bg_n_range = (5, 25),
# int_range = (1., 1.),
# max_rotation = 45,
# is_v_flip = False)
#argkws = dict(output_size = 256)
gen = MNISTFashionFlow(is_separate=True, fg_n_range=(1, 5), **argkws)
gen.test()
Created on Fri Aug 17 15:08:14 2018
@author: avelinojaver
"""
import sys
from pathlib import Path
dname = Path(__file__).resolve().parents[1]
sys.path.append(str(dname))
from noise2noise.flow import CroppedFlow
from noise2noise.models import UNet
from torch import nn
if __name__ == '__main__':
from torch.utils.data import DataLoader
import tqdm
gen = CroppedFlow()
loader = DataLoader(gen, batch_size=8)
gen.train()
tops = []
bots = []
mod = UNet(n_channels = 1, n_classes = 1)
for X,Y in tqdm.tqdm(loader):
Xhat = mod(X)
break
from noise2noise.models import UNet
from noise2noise.trainer import log_dir_root
from read_movies.moviereader import MovieReader
import numpy as np
import torch
import tqdm
import math
from scipy.ndimage.filters import median_filter
import torch.nn.functional as F
if __name__ == '__main__':
model = UNet(n_channels=1, n_classes=1)
#model_path = log_dir_root / 'l2_20180819_122435_unet_adam_lr0.0001_wd0.0_batch8' / 'checkpoint.pth.tar'
model_path = log_dir_root / 'l1_20180819_122435_unet_adam_lr0.0001_wd0.0_batch8' / 'checkpoint.pth.tar'
state = torch.load(model_path, map_location='cpu')
model.load_state_dict(state['state_dict'])
#movie_name = Path.home() / 'workspace/Vesicles/data/22_09_16/ves5/ramp100.22Sep2016_17.49.11.movie'
#frame_number = 200
#movie_name = '/Users/avelinojaver/OneDrive - Nexus365/vesicle/data/script_ramp.08Dec2015_17.09.35.movie'
movie_name = '/Users/avelinojaver/OneDrive - Nexus365/vesicle/data/script_ramp.08Dec2015_16.45.56.movie'
frame_number = 1066 #tot-1#1761
reader = MovieReader(str(movie_name))
block[ii - xhat.shape[0] + 1:ii + 1] = xhat
if __name__ == '__main__':
cuda_id = 0
batch_size = 2
scale_log = (7, 11.1)
#root_dir = Path('/Users/avelinojaver/OneDrive - Nexus365/microglia/hdf_movies/movies/2018.08.22_movies/180822_MicVid_20X_Dispense/180822_MicVid_20X_Dispense_J9-Media_J11-100uMATP_1/')
save_dir = Path.home() / 'workspace/Vesicles/movies_cleaned/'
root_dir = Path.home() / 'workspace/Vesicles/data/'
model_path = log_dir_root / 'l1_20180819_122435_unet_adam_lr0.0001_wd0.0_batch8' / 'checkpoint.pth.tar'
model = UNet(n_channels=1, n_classes=1)
state = torch.load(str(model_path), map_location='cpu')
model.load_state_dict(state['state_dict'])
if torch.cuda.is_available():
print("THIS IS CUDA!!!!")
dev_str = "cuda:" + str(cuda_id)
else:
dev_str = 'cpu'
device = torch.device(dev_str)
model = model.to(device)
model.eval()
fnames = list(root_dir.rglob('*.movie'))