def load_pointnet(model_path):
model = Pointnet(num_classes=4, input_channels=72, use_xyz=False)
if os.path.isfile(model_path):
print("==> Loading from checkpoint '{}'".format(model_path))
checkpoint = torch.load(model_path)
epoch = checkpoint["epoch"]
it = checkpoint.get("it", 0.0)
best_prec = checkpoint["best_prec"]
if model is not None and checkpoint["model_state"] is not None:
model.load_state_dict(checkpoint["model_state"])
print("==> Done")
return model
test_set = Kitti3DSemSeg(args.num_points, train=False)
test_loader = DataLoader(test_set,
batch_size=args.batch_size,
shuffle=True,
pin_memory=True,
num_workers=0)
# train_set = Indoor3DSemSeg(args.num_points)
train_set = Kitti3DSemSeg(args.num_points)
train_loader = DataLoader(train_set,
batch_size=args.batch_size,
pin_memory=True,
num_workers=0,
shuffle=True)
model = Pointnet(num_classes=34, input_channels=6, use_xyz=True)
# 设置设备
device_name = "cpu"
if torch.cuda.is_available():
device_name = "cuda"
torch.backends.cudnn.deterministic = True
torch.cuda.manual_seed(0)
device = torch.device(device_name)
print(device)
model.cuda()
optimizer = optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
lr_lbmd = lambda it: max(
batch_size=args.batch_size,
shuffle=True,
num_workers=4,
pin_memory=True,
)
train_set = ModelNet40Cls(args.num_points, transforms=train_transforms)
train_loader = DataLoader(
train_set,
batch_size=args.batch_size,
shuffle=True,
num_workers=4,
pin_memory=True,
)
model = Pointnet(input_channels=0, num_classes=40, use_xyz=True)
model.cuda()
optimizer = optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
lr_lbmd = lambda it: max(
args.lr_decay**(int(it * args.batch_size / args.decay_step)),
lr_clip / args.lr,
)
bn_lbmd = lambda it: max(
args.bn_momentum * args.bnm_decay**
(int(it * args.batch_size / args.decay_step)),
bnm_clip,
)
train_dataset = TripletFaceDataset(root='',
n_triplets=args.num_triplet,
n_points=args.num_points,
class_nums=500,
transforms=None,
extensions='bcnc')
train_loader = DataLoader(
train_dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=4,
pin_memory=True,
)
model = Pointnet(input_channels=3, use_xyz=True)
model.cuda()
# 512 is dimension of feature
classifier = {
'MCP': layer.MarginCosineProduct(512, args.num_class).cuda(),
'AL': layer.AngleLinear(512, args.num_class).cuda(),
'L': torch.nn.Linear(512, args.num_class, bias=False).cuda()
}[args.classifier_type]
criterion = nn.TripletMarginLoss(margin=0.5, p=2)
optimizer = optim.Adam([{
'params': model.parameters()
}, {
'params': classifier.parameters()
}],
lr=lr,
test_sampler = SubsetRandomSampler(val_indices)
test_loader = DataLoader(dataset,
batch_size=args.batch_size,
shuffle=False,
pin_memory=True,
num_workers=4,
sampler=test_sampler)
train_loader = DataLoader(dataset,
batch_size=args.batch_size,
shuffle=False,
pin_memory=True,
num_workers=4,
sampler=train_sampler)
model = Pointnet(input_channels=1, num_classes=8, use_xyz=True)
model.cuda()
optimizer = optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
def lr_lbmd(it):
return max(
args.lr_decay**(int(it * args.batch_size / args.decay_step)),
lr_clip / args.lr,
)
def bn_lbmd(it):
return max(
args.bn_momentum *
args.bnm_decay**(int(it * args.batch_size / args.decay_step)),
shuffle=True,
num_workers=2,
pin_memory=True,
)
train_set = ModelNet40Cls(args.num_points, transforms=transforms)
train_loader = DataLoader(
train_set,
batch_size=args.batch_size,
shuffle=True,
num_workers=2,
pin_memory=True,
)
model = Pointnet(input_channels=0,
num_classes=40,
use_xyz=False,
type_r=args.typer)
model.cuda()
optimizer = optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
lr_lbmd = lambda it: max(
args.lr_decay**(int(it * args.batch_size / args.decay_step)),
lr_clip / args.lr,
)
bn_lbmd = lambda it: max(
args.bn_momentum * args.bnm_decay**
(int(it * args.batch_size / args.decay_step)),
bnm_clip,
)
batch_size=args.batch_size,
pin_memory=True,
num_workers=2,
shuffle=True,
)
train_set = Indoor3DSemSeg(args.num_points,use_colour=args.colours,split="train")
train_loader = DataLoader(
train_set,
batch_size=args.batch_size,
pin_memory=True,
num_workers=2,
shuffle=True,
)
if args.colours:
model = Pointnet(num_classes=27, input_channels=7, use_xyz=True)
else:
model = Pointnet(num_classes=27, input_channels=3, use_xyz=True)
model.cuda()
optimizer = optim.Adam(
model.parameters(), lr=args.lr, weight_decay=args.weight_decay
)
lr_lbmd = lambda it: max(
args.lr_decay ** (int(it * args.batch_size / args.decay_step)),
lr_clip / args.lr,
)
bnm_lmbd = lambda it: max(
args.bn_momentum
* args.bn_decay ** (int(it * args.batch_size / args.decay_step)),
bnm_clip,
batch_size=4096,
shuffle=True,
pin_memory=True,
num_workers=2,
)
train_set = RadarLowLvlSemSeg(args.num_points, train_rat=0.6)
train_loader = DataLoader(
train_set,
batch_size=args.batch_size,
pin_memory=True,
num_workers=2,
shuffle=True,
)
model = Pointnet(num_classes=4, input_channels=72, use_xyz=False)
model.cuda()
optimizer = optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
lr_lbmd = lambda it: max(
args.lr_decay**(int(it * args.batch_size / args.decay_step)),
lr_clip / args.lr,
)
bnm_lmbd = lambda it: max(
args.bn_momentum * args.bn_decay**
(int(it * args.batch_size / args.decay_step)),
bnm_clip,
)
shuffle=True,
num_workers=2,
pin_memory=True,
drop_last=True)
val_loader = DataLoader(val_set,
batch_size=args.batch_size,
shuffle=False,
num_workers=2,
pin_memory=True,
drop_last=False)
# model
device = torch.device("cuda" if args.cuda else "cpu")
model = Pointnet(input_channels=args.input_plane,
num_classes=args.output_plane,
use_xyz=True)
model.to(device)
model = nn.DataParallel(model)
print("Let's use", torch.cuda.device_count(), "GPUs!")
# optimizer
optimizer = optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
# scheduler
lr_lbmd = lambda it: max(
args.lr_decay**(int(it * args.batch_size / args.decay_step)),
lr_clip / args.lr,
)
division,
absolute_import,
with_statement,
print_function,
unicode_literals,
)
import numpy as np
import time
import glob
import torch
import torch.optim as optim
import torch.optim.lr_scheduler as lr_sched
import torch.nn as nn
from torch.utils.data import DataLoader
from torchvision import transforms
import etw_pytorch_utils as pt_utils
import pprint
import os.path as osp
import os, sys
import argparse
sys.path.append('../..')
from pointnet2.models import Pointnet2ClsMSG as Pointnet
from pointnet2.models.pointnet2_msg_cls import model_fn_decorator
from pointnet2.data import ScanNet
import pointnet2.data.data_utils as d_utils
if __name__ == "__main__":
# model
model = Pointnet(input_channels=3, num_classes=21, use_xyz=True)
print('#parameters %d' % sum([x.nelement() for x in model.parameters()]))
batch_size=args.batch_size,
shuffle=True,
pin_memory=True,
num_workers=2,
)
print("Loaded Test Set")
train_set = Indoor3DSemSeg(args.num_points)
train_loader = DataLoader(
train_set,
batch_size=args.batch_size,
pin_memory=True,
num_workers=2,
shuffle=True,
)
print("Loaded Train Set")
model = Pointnet(num_classes=21, input_channels=0, use_xyz=True)
model.cuda()
print("Initialized Model")
optimizer = optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
lr_lbmd = lambda it: max(
args.lr_decay**(int(it * args.batch_size / args.decay_step)),
lr_clip / args.lr,
)
bnm_lmbd = lambda it: max(
args.bn_momentum * args.bn_decay**
(int(it * args.batch_size / args.decay_step)),
bnm_clip,
)
gpu_index = paras.gpu
os.environ["CUDA_VISIBLE_DEVICES"] = gpu_index
lr_clip = 1e-5
bnm_clip = 1e-2
if __name__ == "__main__":
args = parser.parse_args()
test_set = Indoor3DSemSeg(args.num_points, train=False)
test_loader = DataLoader(test_set,
batch_size=args.batch_size,
shuffle=False,
pin_memory=True,
num_workers=2)
model = Pointnet(3, input_channels=6).cuda()
model.load_state_dict(
torch.load(
'/home1/xuhui/Pointnet2_PyTorch-master/pointnet2/train/checkpoints/poitnet2_semseg_best.pth.tar'
)['model_state'])
model.eval()
# for i, data in tqdm.tqdm(enumerate(test_loader, 0), total=len(test_loader),
# leave=False, desc='val'):
for i, data in enumerate(test_loader):
f = open(os.path.join(preds_path, str(i).zfill(2) + '.txt'), 'w')
inputs, labels = data
inputs = inputs.to('cuda', non_blocking=True)
labels = labels.to('cuda', non_blocking=True)
# print(labels)
preds = model(inputs)
