import configs.config_loader as cfg_loader
import os
import trimesh
import numpy as np
from tqdm import tqdm
cfg = cfg_loader.get_config()
device = torch.device("cuda")
net = model.NDF()
dataset = voxelized_data.VoxelizedDataset('test',
res=cfg.input_res,
pointcloud_samples=cfg.num_points,
data_path=cfg.data_dir,
split_file=cfg.split_file,
batch_size=1,
num_sample_points=cfg.num_sample_points_generation,
num_workers=30,
sample_distribution=cfg.sample_ratio,
sample_sigmas=cfg.sample_std_dev)
gen = Generator(net, cfg.exp_name, device=device)
out_path = 'experiments/{}/evaluation/'.format(cfg.exp_name)
def gen_iterator(out_path, dataset, gen_p):
global gen
gen = gen_p
if not os.path.exists(out_path):
args = parser.parse_known_args()[0]
if args.model == 'ShapeNet32Vox':
net = model.ShapeNet32Vox()
if args.model == 'ShapeNet128Vox':
net = model.ShapeNet128Vox()
if args.model == 'ShapeNetPoints':
net = model.ShapeNetPoints()
train_dataset = voxelized_data.VoxelizedDataset(
'train',
voxelized_pointcloud=args.pointcloud,
pointcloud_samples=args.pc_samples,
res=args.res,
sample_distribution=args.sample_distribution,
sample_sigmas=args.sample_sigmas,
num_sample_points=50000,
batch_size=args.batch_size,
num_workers=30)
val_dataset = voxelized_data.VoxelizedDataset(
'val',
voxelized_pointcloud=args.pointcloud,
pointcloud_samples=args.pc_samples,
res=args.res,
sample_distribution=args.sample_distribution,
sample_sigmas=args.sample_sigmas,
num_sample_points=50000,
batch_size=args.batch_size,
num_workers=30)
import models.data.voxelized_data_shapenet as voxelized_data
from models import training
import torch
import configs.config_loader as cfg_loader
cfg = cfg_loader.get_config()
net = model.NDF()
train_dataset = voxelized_data.VoxelizedDataset('train',
res=cfg.input_res,
pointcloud_samples=cfg.num_points,
data_path=cfg.data_dir,
split_file=cfg.split_file,
batch_size=cfg.batch_size,
num_sample_points=cfg.num_sample_points_training,
num_workers=30,
sample_distribution=cfg.sample_ratio,
sample_sigmas=cfg.sample_std_dev)
val_dataset = voxelized_data.VoxelizedDataset('val',
res=cfg.input_res,
pointcloud_samples=cfg.num_points,
data_path=cfg.data_dir,
split_file=cfg.split_file,
batch_size=cfg.batch_size,
num_sample_points=cfg.num_sample_points_training,
num_workers=30,
sample_distribution=cfg.sample_ratio,
sample_sigmas=cfg.sample_std_dev)
net = model.ShapeNet32Vox()
if args.model == 'ShapeNet128Vox':
net = model.ShapeNet128Vox()
if args.model == 'ShapeNetPoints':
net = model.ShapeNetPoints()
if args.model == 'SVR':
net = model.SVR()
dataset = voxelized_data.VoxelizedDataset(
args.mode,
voxelized_pointcloud=args.pointcloud,
pointcloud_samples=args.pc_samples,
res=args.res,
sample_distribution=args.sample_distribution,
sample_sigmas=args.sample_sigmas,
num_sample_points=100,
batch_size=1,
num_workers=0)
exp_name = 'i{}_dist-{}sigmas-{}v{}_m{}'.format(
'PC' + str(args.pc_samples) if args.pointcloud else 'Voxels',
''.join(str(e) + '_' for e in args.sample_distribution),
''.join(str(e) + '_' for e in args.sample_sigmas), args.res, args.model)
gen = Generator(net,
0.5,
exp_name,
checkpoint=args.checkpoint,
resolution=args.retrieval_res,
def loadNDF(index,
pointcloud_samples,
exp_name,
data_dir,
split_file,
sample_distribution,
sample_sigmas,
res,
mode='test'):
global encoding
global net
global device
net = model.NDF()
device = torch.device("cuda")
if 'garments' in exp_name.lower():
dataset = dataloader_garments.VoxelizedDataset(
mode=mode,
data_path=data_dir,
split_file=split_file,
res=res,
density=0,
pointcloud_samples=pointcloud_samples,
sample_distribution=sample_distribution,
sample_sigmas=sample_sigmas,
)
checkpoint = 'checkpoint_127h:6m:33s_457593.9149734974'
generator = generation.Generator(net,
exp_name,
checkpoint=checkpoint,
device=device)
if 'cars' in exp_name.lower():
dataset = voxelized_data_shapenet.VoxelizedDataset(
mode=mode,
res=res,
pointcloud_samples=pointcloud_samples,
data_path=data_dir,
split_file=split_file,
sample_distribution=sample_distribution,
sample_sigmas=sample_sigmas,
batch_size=1,
num_sample_points=1024,
num_workers=1)
checkpoint = 'checkpoint_108h:5m:50s_389150.3971107006'
generator = generation.Generator(net,
exp_name,
checkpoint=checkpoint,
device=device)
example = dataset[index]
print('Object: ', example['path'])
inputs = torch.from_numpy(example['inputs']).unsqueeze(0).to(
device) # lead inputs and samples including one batch channel
for param in net.parameters():
param.requires_grad = False
encoding = net.encoder(inputs)