def gen_objective_str(targetid, model, params, neuron=False):
print('generating feature_viz objective string for %s' % targetid)
if '-' in str(targetid):
within_id = edgename_2_singlenum(model, targetid, params)
layer_name = nodeid_2_perlayerid(targetid.split('-')[1], params)[2]
if not neuron:
return layer_name + '_preadd_conv:' + str(within_id)
else:
return objectives.neuron(layer_name + '_preadd_conv',
int(within_id))
else:
layer, within_id, layer_name = nodeid_2_perlayerid(targetid, params)
if not neuron:
return layer_name + ':' + str(within_id)
else:
return objectives.neuron(layer_name, int(within_id))
def test_gan_img_optim(inceptionv1_model):
""" Test if GAN generated image could be optimized """
G = upconvGAN("fc6").to(device)
def GANparam(batch=1, sd=1):
code = (torch.randn(
(batch, G.codelen)) * sd).to(device).requires_grad_(True)
imagef = lambda: G.visualize(code)
return [code], imagef
objective = objectives.neuron("input", 0)
assert_gan_gradient_descent(GANparam, objective, inceptionv1_model)
def test_neuron(inceptionv1_model):
objective = objectives.neuron("mixed3a_1x1_pre_relu_conv", 0)
assert_gradient_descent(objective, inceptionv1_model)
def fetch_deepviz_img_for_subgraph(model, layer_name, within_id, targetid,
viz_folder, params):
model = set_across_model(model, 'target_node', None)
objective_str = layer_name + ':' + str(targetid)
neuron = params['deepviz_neuron']
#generate objective
#objective = gen_objective(targetid,model,params,neuron=neuron)
print('generating feature_viz objective string for %s' % targetid)
if not params['deepviz_neuron']:
#return layer_name+':'+str(within_id)
objective = objectives.channel(layer_name, int(within_id))
else:
objective = objectives.neuron(layer_name, int(within_id))
file_path = viz_folder + '/images.csv'
parametrizer = params['deepviz_param']
optimizer = params['deepviz_optim']
transforms = params['deepviz_transforms']
image_size = params['deepviz_image_size']
param_str = object_2_str(parametrizer, "params['deepviz_param']=")
optimizer_str = object_2_str(optimizer, "params['deepviz_optim']=")
transforms_str = object_2_str(transforms, "params['deepviz_transforms']=")
df = pd.read_csv(file_path, dtype=str)
df_sel = df.loc[(df['targetid'] == str(targetid))
& (df['objective'] == objective_str) &
(df['parametrizer'] == param_str) &
(df['optimizer'] == optimizer_str) &
(df['transforms'] == transforms_str) &
(df['neuron'] == str(neuron))]
if len(df_sel) == 0:
print('deepviz image not found for %s, generating . . .' % targetid)
#image_name = 'deepviz_'+str(targetid)+'_'+objective+'_'+str(time.time())+'.jpg'
image_name = str(targetid) + '_' + objective_str + '_' + str(
time.time()) + '.jpg'
#gen_visualization(model,image_name,objective,parametrizer,optimizer,transforms,image_size,neuron,params)
if neuron:
full_image_path = viz_folder + '/neuron/' + image_name
else:
full_image_path = viz_folder + '/channel/' + image_name
if parametrizer is None:
parametrizer = lambda: param.image(image_size)
print('generating featviz with objective: %s' % str(objective_str))
_ = render.render_vis(model,
objective,
parametrizer,
optimizer,
transforms=transforms,
save_image=True,
image_name=full_image_path,
show_inline=True)
with open(file_path, 'a') as csv:
csv.write(','.join([
image_name,
str(targetid), objective_str, param_str, optimizer_str,
transforms_str,
str(neuron)
]) + '\n')
else:
print('found pre-generated image')
image_name = df_sel.iloc[0]['image_name']
if neuron:
return 'neuron/' + image_name
else:
return 'channel/' + image_name