def define_model(self):
opts = self.opts
# ----------
# Options
# ----------
self.symmetric = opts.symmetric
anno_sfm_path = osp.join(opts.cub_cache_dir, 'sfm', 'anno_train.mat')
anno_sfm = sio.loadmat(anno_sfm_path,
struct_as_record=False,
squeeze_me=True)
if opts.sphere_initial:
sfm_mean_shape = mesh.create_sphere(3)
else:
sfm_mean_shape = (np.transpose(anno_sfm['S']),
anno_sfm['conv_tri'] - 1)
img_size = (opts.img_size, opts.img_size)
self.model = mesh_net.MeshNet(img_size,
opts,
nz_feat=opts.nz_feat,
num_kps=opts.num_kps,
sfm_mean_shape=sfm_mean_shape)
if opts.num_pretrain_epochs > 0:
self.load_network(self.model, 'pred', opts.num_pretrain_epochs)
self.model = self.model.cuda(device=opts.gpu_id)
# Data structures to use for triangle priors.
edges2verts = self.model.edges2verts
# B x E x 4
edges2verts = np.tile(np.expand_dims(edges2verts, 0),
(opts.batch_size, 1, 1))
self.edges2verts = Variable(
torch.LongTensor(edges2verts).cuda(device=opts.gpu_id),
requires_grad=False)
# For renderering.
faces = self.model.faces.view(1, -1, 3)
self.faces = faces.repeat(opts.batch_size, 1, 1)
# opts.renderer = "smr"
self.renderer = NeuralRenderer(
opts.img_size) if opts.renderer == "nmr" else SoftRenderer(
opts.img_size)
self.renderer_predcam = NeuralRenderer(
opts.img_size) if opts.renderer == "nmr" else SoftRenderer(
opts.img_size) #for camera loss via projection
# Need separate NMR for each fwd/bwd call.
if opts.texture:
self.tex_renderer = NeuralRenderer(
opts.img_size) if opts.renderer == "nmr" else SoftRenderer(
opts.img_size)
# Only use ambient light for tex renderer
self.tex_renderer.ambient_light_only()
# For visualization
self.vis_rend = bird_vis.VisRenderer(opts.img_size,
faces.data.cpu().numpy())
return
def __init__(self, opts):
self.opts = opts
self.symmetric = opts.symmetric
img_size = (opts.img_size, opts.img_size)
print('Setting up model..')
self.model = mesh_net.MeshNet(img_size, opts, nz_feat=opts.nz_feat)
self.load_network(self.model, 'pred', self.opts.num_train_epoch)
self.model.eval()
self.model = self.model.cuda(device=self.opts.gpu_id)
# TODO junzhe option of renderer
print('self.opts.renderer_opt:', self.opts.renderer_opt)
if self.opts.renderer_opt == 'nmr':
from nnutils.nmr import NeuralRenderer
elif self.opts.renderer_opt == 'nmr_kaolin':
from nnutils.nmr_kaolin import NeuralRenderer
elif self.opts.renderer_opt == 'dibr_kaolin':
from nnutils.dibr_kaolin import NeuralRenderer
else:
raise NotImplementedError
self.renderer = NeuralRenderer(opts.img_size,
uv_sampler=self.model.uv_sampler)
if opts.texture:
self.tex_renderer = NeuralRenderer(
opts.img_size, uv_sampler=self.model.uv_sampler)
# Only use ambient light for tex renderer
self.tex_renderer.ambient_light_only()
if opts.use_sfm_ms:
anno_sfm_path = osp.join(opts.cub_cache_dir, 'sfm',
'anno_testval.mat')
anno_sfm = sio.loadmat(anno_sfm_path,
struct_as_record=False,
squeeze_me=True)
sfm_mean_shape = torch.Tensor(np.transpose(
anno_sfm['S'])).cuda(device=opts.gpu_id)
self.sfm_mean_shape = Variable(sfm_mean_shape, requires_grad=False)
self.sfm_mean_shape = self.sfm_mean_shape.unsqueeze(0).repeat(
opts.batch_size, 1, 1)
sfm_face = torch.LongTensor(anno_sfm['conv_tri'] -
1).cuda(device=opts.gpu_id)
self.sfm_face = Variable(sfm_face, requires_grad=False)
faces = self.sfm_face.view(1, -1, 3)
else:
# For visualization
faces = self.model.faces.view(1, -1, 3)
self.faces = faces.repeat(opts.batch_size, 1, 1)
self.vis_rend = bird_vis.VisRenderer(opts.img_size,
faces.data.cpu().numpy(),
self.opts,
uv_sampler=self.model.uv_sampler)
self.vis_rend.set_bgcolor([1., 1., 1.])
self.resnet_transform = torchvision.transforms.Normalize(
mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
def define_model(self):
# 2
opts = self.opts
# import pdb; pdb.set_trace()
# Options
self.symmetric = opts.symmetric
anno_sfm_path = os.path.join(opts.cub_cache_dir, 'sfm',
'anno_train.mat')
anno_sfm = sio.loadmat(anno_sfm_path,
struct_as_record=False,
squeeze_me=True)
sfm_mean_shape = (np.transpose(anno_sfm['S']),
anno_sfm['conv_tri'] - 1)
img_size = (opts.img_size, opts.img_size)
self.model = mesh_net.MeshNet(img_size,
opts,
nz_feat=opts.nz_feat,
num_kps=opts.num_kps,
sfm_mean_shape=sfm_mean_shape)
if opts.num_pretrain_epochs > 0:
self.load_network(self.model, 'pred', opts.num_pretrain_epochs)
self.model = self.model.to(self.device)
# Data structures to use for triangle priors.
edges2verts = self.model.edges2verts
# B x E x 4
edges2verts = np.tile(np.expand_dims(edges2verts, 0),
(opts.batch_size, 1, 1))
self.edges2verts = Variable(
torch.LongTensor(edges2verts).cuda(device=self.device),
requires_grad=False)
# For renderering.
faces = self.model.faces.view(1, -1, 3)
self.faces = faces.repeat(opts.batch_size, 1, 1)
self.renderer = NeuralRenderer(opts.img_size)
self.renderer_predcam = NeuralRenderer(
opts.img_size) #for camera loss via projection
# Need separate NMR for each fwd/bwd call.
if opts.texture:
self.tex_renderer = NeuralRenderer(opts.img_size)
# Only use ambient light for tex renderer
self.tex_renderer.ambient_light_only()
# For visualization
self.vis_rend = bird_vis.VisRenderer(opts.img_size,
faces.data.cpu().numpy())
# import ipdb
# ipdb.set_trace()
# for k,v in self.model.named_modules():
# v.register_backward_hook(hook)
return
def __init__(self, opts):
self.opts = opts
self.symmetric = opts.symmetric
#img_size是(256,256)
img_size = (opts.img_size, opts.img_size)
print('Setting up model..')
#-----------------目前猜測是在這一行的什後從mean mesh變成learned mesh的
# print(opts.nz_feat)
# exit()
#nz_feat目前不確定是哪冒出來的，還要找源頭
#nz_feat 為200
self.model = mesh_net.MeshNet(img_size, opts, nz_feat=opts.nz_feat)
#-----------------------------------經這一個之後就被改變了得到一個337的verts,但原本的verts至少有600個所以它可能是將某些點更動了,
# 也可能是它會透過對稱的手法來變成完整的mean shape
self.load_network(self.model, 'pred', self.opts.num_train_epoch)
#model 從training()模式轉換成評估模式
self.model.eval()
self.model = self.model.cuda(device=self.opts.gpu_id)
self.renderer = NeuralRenderer(opts.img_size)
if opts.texture: #--------------------這個只是true而已
self.tex_renderer = NeuralRenderer(opts.img_size)
# Only use ambient light for tex renderer
self.tex_renderer.ambient_light_only()
#--------------------------------這邊將initial mean shape拿進去訓練得到 訓練過後的learned mean shape
#----------------是否使用use_sfm_ms(它門預設都沒有，這個mesh非常的簡陋，它必須經過學習才會得到一個mean shape
if opts.use_sfm_ms:
anno_sfm_path = osp.join(opts.cub_cache_dir, 'sfm',
'anno_testval.mat')
anno_sfm = sio.loadmat(anno_sfm_path,
struct_as_record=False,
squeeze_me=True)
sfm_mean_shape = torch.Tensor(np.transpose(
anno_sfm['S'])).cuda(device=opts.gpu_id)
self.sfm_mean_shape = Variable(sfm_mean_shape, requires_grad=False)
self.sfm_mean_shape = self.sfm_mean_shape.unsqueeze(0).repeat(
opts.batch_size, 1, 1)
sfm_face = torch.LongTensor(anno_sfm['conv_tri'] -
1).cuda(device=opts.gpu_id)
self.sfm_face = Variable(sfm_face, requires_grad=False)
faces = self.sfm_face.view(1, -1, 3)
#-------------------------------------------
else:
# For visualization
faces = self.model.faces.view(1, -1, 3)
self.faces = faces.repeat(opts.batch_size, 1, 1)
#--------------------------------------這邊會到vis render init()
self.vis_rend = bird_vis.VisRenderer(opts.img_size,
faces.data.cpu().numpy())
self.vis_rend.set_bgcolor([1., 1., 1.])
self.resnet_transform = torchvision.transforms.Normalize(
mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
def define_model(self):
opts = self.opts
# ----------
# Options
# ----------
#是否對稱
self.symmetric = opts.symmetric
#--------------parker 不確定這是幹麻的
#--------------anno_sfm_path會得到一個路徑
anno_sfm_path = osp.join(opts.cub_cache_dir, 'sfm', 'anno_train.mat')
#----anno_train.mat是只有15個點的最基礎的鳥模型
anno_sfm = sio.loadmat(anno_sfm_path,
struct_as_record=False,
squeeze_me=True)
#----------將anno_sfm["S"]是15個點的vertex，anno_sfm["conv_tri"]是15個點組成的面
sfm_mean_shape = (np.transpose(anno_sfm['S']),
anno_sfm['conv_tri'] - 1)
#--------------
img_size = (opts.img_size, opts.img_size)
#這邊會進去mesh_net.py的MeshNet
self.model = mesh_net.MeshNet(img_size,
opts,
nz_feat=opts.nz_feat,
num_kps=opts.num_kps,
sfm_mean_shape=sfm_mean_shape)
#-----如果有已經訓練過得epochs則執行這一行
if opts.num_pretrain_epochs > 0:
self.load_network(self.model, 'pred', opts.num_pretrain_epochs)
self.model = self.model.cuda(device=opts.gpu_id)
# Data structures to use for triangle priors.
#---------這邊是拿取modle已經計算好的eges2verts
edges2verts = self.model.edges2verts
# B x E x 4
edges2verts = np.tile(np.expand_dims(edges2verts, 0),
(opts.batch_size, 1, 1))
self.edges2verts = Variable(
torch.LongTensor(edges2verts).cuda(device=opts.gpu_id),
requires_grad=False)
# For renderering.
faces = self.model.faces.view(1, -1, 3)
self.faces = faces.repeat(opts.batch_size, 1, 1)
#include nmr並且取名叫做Neural Renderer
self.renderer = NeuralRenderer(opts.img_size)
self.renderer_predcam = NeuralRenderer(
opts.img_size) #for camera loss via projection
#如果要計算texture的話，會執行下面這一行，但我反而不懂，為什麼要執行這一行
# Need separate NMR for each fwd/bwd call.
if opts.texture:
self.tex_renderer = NeuralRenderer(opts.img_size)
# Only use ambient light for tex renderer
self.tex_renderer.ambient_light_only()
# For visualization
self.vis_rend = bird_vis.VisRenderer(opts.img_size,
faces.data.cpu().numpy())
# import ipdb
# ipdb.set_trace()
# for k,v in self.model.named_modules():
# v.register_backward_hook(hook)
return