From 0be0ecf0a8fc6df1f9e354f8aea12b7008f658f1 Mon Sep 17 00:00:00 2001
From: natanielruiz <nataniel777@hotmail.com>
Date: 星期三, 27 九月 2017 06:21:54 +0800
Subject: [PATCH] hopenet experiments
---
code/hopenet.py | 90 ++++++++
code/train_finetune_new.py | 192 +++++++++++++++++
code/test_new.py | 136 ++++++++++++
code/train_hopenet_new.py | 221 ++++++++++++++++++++
4 files changed, 639 insertions(+), 0 deletions(-)
diff --git a/code/hopenet.py b/code/hopenet.py
index c6bf0db..de2f4ec 100644
--- a/code/hopenet.py
+++ b/code/hopenet.py
@@ -340,3 +340,93 @@
angles.append(preangles)
return pre_yaw, pre_pitch, pre_roll, angles, sr_output
+
+class Hopenet_new(nn.Module):
+ # This is just Hopenet with 3 output layers for yaw, pitch and roll.
+ def __init__(self, block, layers, num_bins):
+ self.inplanes = 64
+ super(Hopenet_new, self).__init__()
+ self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,
+ bias=False)
+ self.bn1 = nn.BatchNorm2d(64)
+ self.relu = nn.ReLU(inplace=True)
+ self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
+ self.layer1 = self._make_layer(block, 64, layers[0])
+ self.layer2 = self._make_layer(block, 128, layers[1], stride=2)
+ self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
+ self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
+ self.avgpool = nn.AvgPool2d(7)
+ self.fc_yaw = nn.Linear(512 * block.expansion, num_bins)
+ self.fc_pitch = nn.Linear(512 * block.expansion, num_bins)
+ self.fc_roll = nn.Linear(512 * block.expansion, num_bins)
+
+ self.softmax = nn.Softmax()
+ self.fc_finetune_new = nn.Linear(512 * block.expansion + 256 * block.expansion + 3, 3)
+ self.conv1x1 = nn.Conv2d(1024, 64, kernel_size = 1, stride = 1, bias=False)
+ self.maxpool_interm = nn.MaxPool2d(kernel_size=5, stride=3, padding=1)
+
+ self.idx_tensor = Variable(torch.FloatTensor(range(66))).cuda()
+
+ for m in self.modules():
+ if isinstance(m, nn.Conv2d):
+ n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
+ m.weight.data.normal_(0, math.sqrt(2. / n))
+ elif isinstance(m, nn.BatchNorm2d):
+ m.weight.data.fill_(1)
+ m.bias.data.zero_()
+
+ def _make_layer(self, block, planes, blocks, stride=1):
+ downsample = None
+ if stride != 1 or self.inplanes != planes * block.expansion:
+ downsample = nn.Sequential(
+ nn.Conv2d(self.inplanes, planes * block.expansion,
+ kernel_size=1, stride=stride, bias=False),
+ nn.BatchNorm2d(planes * block.expansion),
+ )
+
+ layers = []
+ layers.append(block(self.inplanes, planes, stride, downsample))
+ self.inplanes = planes * block.expansion
+ for i in range(1, blocks):
+ layers.append(block(self.inplanes, planes))
+
+ return nn.Sequential(*layers)
+
+ def forward(self, x):
+ x = self.conv1(x)
+ x = self.bn1(x)
+ x = self.relu(x)
+ x = self.maxpool(x)
+
+ x = self.layer1(x)
+ x = self.layer2(x)
+ x = self.layer3(x)
+ x_interm = self.conv1x1(x)
+ x_interm = self.relu(x_interm)
+ x_interm = self.maxpool_interm(x_interm)
+ x_interm = x_interm.view(x_interm.size(0), -1)
+
+ x = self.layer4(x)
+
+ x = self.avgpool(x)
+ x = x.view(x.size(0), -1)
+ pre_yaw = self.fc_yaw(x)
+ pre_pitch = self.fc_pitch(x)
+ pre_roll = self.fc_roll(x)
+
+ yaw = self.softmax(pre_yaw)
+ yaw = Variable(torch.sum(yaw.data * self.idx_tensor.data, 1), requires_grad=True) * 3 - 99
+ pitch = self.softmax(pre_pitch)
+ pitch = Variable(torch.sum(pitch.data * self.idx_tensor.data, 1), requires_grad=True) * 3 - 99
+ roll = self.softmax(pre_roll)
+ roll = Variable(torch.sum(roll.data * self.idx_tensor.data, 1), requires_grad=True) * 3 - 99
+ yaw = yaw.view(yaw.size(0), 1)
+ pitch = pitch.view(pitch.size(0), 1)
+ roll = roll.view(roll.size(0), 1)
+ preangles = torch.cat([yaw, pitch, roll], 1)
+
+ # angles predicts the residual
+ residuals = self.fc_finetune_new(torch.cat((preangles, x_interm, x), 1))
+ final_angles = preangles + residuals
+
+ return pre_yaw, pre_pitch, pre_roll, preangles, final_angles
diff --git a/code/test_new.py b/code/test_new.py
new file mode 100644
index 0000000..3c3f394
--- /dev/null
+++ b/code/test_new.py
@@ -0,0 +1,136 @@
+import numpy as np
+import torch
+import torch.nn as nn
+from torch.autograd import Variable
+from torch.utils.data import DataLoader
+from torchvision import transforms
+import torch.backends.cudnn as cudnn
+import torchvision
+import torch.nn.functional as F
+
+import cv2
+import matplotlib.pyplot as plt
+import sys
+import os
+import argparse
+
+import datasets
+import hopenet
+import utils
+
+def parse_args():
+ """Parse input arguments."""
+ parser = argparse.ArgumentParser(description='Head pose estimation using the Hopenet network.')
+ parser.add_argument('--gpu', dest='gpu_id', help='GPU device id to use [0]',
+ default=0, type=int)
+ parser.add_argument('--data_dir', dest='data_dir', help='Directory path for data.',
+ default='', type=str)
+ parser.add_argument('--filename_list', dest='filename_list', help='Path to text file containing relative paths for every example.',
+ default='', type=str)
+ parser.add_argument('--snapshot', dest='snapshot', help='Path of model snapshot.',
+ default='', type=str)
+ parser.add_argument('--batch_size', dest='batch_size', help='Batch size.',
+ default=1, type=int)
+ parser.add_argument('--save_viz', dest='save_viz', help='Save images with pose cube.',
+ default=False, type=bool)
+ parser.add_argument('--dataset', dest='dataset', help='Dataset type.', default='AFLW2000', type=str)
+
+ args = parser.parse_args()
+
+ return args
+
+if __name__ == '__main__':
+ args = parse_args()
+
+ cudnn.enabled = True
+ gpu = args.gpu_id
+ snapshot_path = args.snapshot
+
+ # ResNet101 with 3 outputs.
+ # model = hopenet.Hopenet(torchvision.models.resnet.Bottleneck, [3, 4, 23, 3], 66)
+ # ResNet50
+ model = hopenet.Hopenet_new(torchvision.models.resnet.Bottleneck, [3, 4, 6, 3], 66)
+ # ResNet18
+ # model = hopenet.Hopenet(torchvision.models.resnet.BasicBlock, [2, 2, 2, 2], 66)
+
+ print 'Loading snapshot.'
+ # Load snapshot
+ saved_state_dict = torch.load(snapshot_path)
+ model.load_state_dict(saved_state_dict)
+
+ print 'Loading data.'
+
+ transformations = transforms.Compose([transforms.Scale(224),
+ transforms.CenterCrop(224), transforms.ToTensor(),
+ transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
+
+ if args.dataset == 'AFLW2000':
+ pose_dataset = datasets.AFLW2000(args.data_dir, args.filename_list,
+ transformations)
+ elif args.dataset == 'BIWI':
+ pose_dataset = datasets.BIWI(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFLW':
+ pose_dataset = datasets.AFLW(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'Pose_300W_LP':
+ pose_dataset = datasets.Pose_300W_LP(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFW':
+ pose_dataset = datasets.AFW(args.data_dir, args.filename_list, transformations)
+ else:
+ print 'Error: not a valid dataset name'
+ sys.exit()
+ test_loader = torch.utils.data.DataLoader(dataset=pose_dataset,
+ batch_size=args.batch_size,
+ num_workers=2)
+
+ model.cuda(gpu)
+
+ print 'Ready to test network.'
+
+ # Test the Model
+ model.eval() # Change model to 'eval' mode (BN uses moving mean/var).
+ total = 0
+ yaw_error = .0
+ pitch_error = .0
+ roll_error = .0
+
+ l1loss = torch.nn.L1Loss(size_average=False)
+
+ for i, (images, labels, cont_labels, name) in enumerate(test_loader):
+ images = Variable(images).cuda(gpu)
+ total += cont_labels.size(0)
+ label_yaw = cont_labels[:,0].float()
+ label_pitch = cont_labels[:,1].float()
+ label_roll = cont_labels[:,2].float()
+
+ pre_yaw, pre_pitch, pre_roll, preangles, final_angles = model(images)
+ yaw = final_angles[:,0].cpu().data
+ pitch = final_angles[:,1].cpu().data
+ roll = final_angles[:,2].cpu().data
+
+ # Mean absolute error
+ yaw_error += torch.sum(torch.abs(yaw - label_yaw))
+ pitch_error += torch.sum(torch.abs(pitch - label_pitch))
+ roll_error += torch.sum(torch.abs(roll - label_roll))
+
+ # Save images with pose cube.
+ # TODO: fix for larger batch size
+ if args.save_viz:
+ name = name[0]
+ if args.dataset == 'BIWI':
+ cv2_img = cv2.imread(os.path.join(args.data_dir, name + '_rgb.png'))
+ else:
+ cv2_img = cv2.imread(os.path.join(args.data_dir, name + '.jpg'))
+
+ if args.batch_size == 1:
+ error_string = 'y %.4f, p %.4f, r %.4f' % (torch.sum(torch.abs(yaw - label_yaw)), torch.sum(torch.abs(pitch - label_pitch)), torch.sum(torch.abs(roll - label_roll)))
+ cv2_img = cv2.putText(cv2_img, error_string, (30, cv2_img.shape[0]- 30), fontFace=1, fontScale=2, color=(0,255,0), thickness=2)
+ utils.plot_pose_cube(cv2_img, yaw[0], pitch[0], roll[0])
+ cv2.imwrite(os.path.join('output/images', name + '.jpg'), cv2_img)
+
+ print('Test error in degrees of the model on the ' + str(total) +
+ ' test images. Yaw: %.4f, Pitch: %.4f, Roll: %.4f' % (yaw_error / total,
+ pitch_error / total, roll_error / total))
+
+ # Binned accuracy
+ # for idx in xrange(len(yaw_correct)):
+ # print yaw_correct[idx] / total, pitch_correct[idx] / total, roll_correct[idx] / total
diff --git a/code/train_finetune_new.py b/code/train_finetune_new.py
new file mode 100644
index 0000000..1f77d54
--- /dev/null
+++ b/code/train_finetune_new.py
@@ -0,0 +1,192 @@
+import numpy as np
+import torch
+import torch.nn as nn
+from torch.autograd import Variable
+from torch.utils.data import DataLoader
+from torchvision import transforms
+import torchvision
+import torch.backends.cudnn as cudnn
+import torch.nn.functional as F
+
+import cv2
+import matplotlib.pyplot as plt
+import sys
+import os
+import argparse
+
+import datasets
+import hopenet
+import torch.utils.model_zoo as model_zoo
+
+model_urls = {
+ 'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',
+ 'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',
+ 'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',
+ 'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',
+ 'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',
+}
+
+def parse_args():
+ """Parse input arguments."""
+ parser = argparse.ArgumentParser(description='Head pose estimation using the Hopenet network.')
+ parser.add_argument('--gpu', dest='gpu_id', help='GPU device id to use [0]',
+ default=0, type=int)
+ parser.add_argument('--num_epochs_ft', dest='num_epochs_ft', help='Maximum number of finetuning epochs.',
+ default=5, type=int)
+ parser.add_argument('--batch_size', dest='batch_size', help='Batch size.',
+ default=16, type=int)
+ parser.add_argument('--lr', dest='lr', help='Base learning rate.',
+ default=0.001, type=float)
+ parser.add_argument('--data_dir', dest='data_dir', help='Directory path for data.',
+ default='', type=str)
+ parser.add_argument('--filename_list', dest='filename_list', help='Path to text file containing relative paths for every example.',
+ default='', type=str)
+ parser.add_argument('--output_string', dest='output_string', help='String appended to output snapshots.', default = '', type=str)
+ parser.add_argument('--alpha', dest='alpha', help='Regression loss coefficient.',
+ default=0.001, type=float)
+ parser.add_argument('--iter_ref', dest='iter_ref', help='Number of iterative refinement passes.',
+ default=1, type=int)
+ parser.add_argument('--dataset', dest='dataset', help='Dataset type.', default='Pose_300W_LP', type=str)
+ parser.add_argument('--snapshot', dest='snapshot', help='Snapshot to start finetuning', default='', type=str)
+ args = parser.parse_args()
+ return args
+
+def get_ignored_params(model):
+ # Generator function that yields ignored params.
+ b = []
+ b.append(model.conv1)
+ b.append(model.bn1)
+ b.append(model.layer1)
+ b.append(model.layer2)
+ b.append(model.layer3)
+ b.append(model.layer4)
+ b.append(model.fc_yaw)
+ b.append(model.fc_pitch)
+ b.append(model.fc_roll)
+ for i in range(len(b)):
+ for module_name, module in b[i].named_modules():
+ if 'bn' in module_name:
+ module.eval()
+ for name, param in module.named_parameters():
+ yield param
+
+def get_non_ignored_params(model):
+ # Generator function that yields params that will be optimized.
+ b = []
+ b.append(model.conv1x1)
+ for i in range(len(b)):
+ for module_name, module in b[i].named_modules():
+ if 'bn' in module_name:
+ module.eval()
+ for name, param in module.named_parameters():
+ yield param
+
+def get_fc_params(model):
+ b = []
+ b.append(model.fc_finetune_new)
+ for i in range(len(b)):
+ for module_name, module in b[i].named_modules():
+ for name, param in module.named_parameters():
+ yield param
+
+def load_filtered_state_dict(model, snapshot):
+ # By user apaszke from discuss.pytorch.org
+ model_dict = model.state_dict()
+ # 1. filter out unnecessary keys
+ snapshot = {k: v for k, v in snapshot.items() if k in model_dict}
+ # 2. overwrite entries in the existing state dict
+ model_dict.update(snapshot)
+ # 3. load the new state dict
+ model.load_state_dict(model_dict)
+
+if __name__ == '__main__':
+ args = parse_args()
+
+ cudnn.enabled = True
+ num_epochs_ft = args.num_epochs_ft
+ batch_size = args.batch_size
+ gpu = args.gpu_id
+
+ if not os.path.exists('output/snapshots'):
+ os.makedirs('output/snapshots')
+
+
+ model = hopenet.Hopenet_new(torchvision.models.resnet.Bottleneck, [3, 4, 6, 3], 66)
+
+ if args.snapshot != '':
+ load_filtered_state_dict(model, torch.load(args.snapshot))
+ else:
+ load_filtered_state_dict(model, model_zoo.load_url(model_urls['resnet50']))
+
+ print 'Loading data.'
+
+ transformations = transforms.Compose([transforms.Scale(240),
+ transforms.RandomCrop(224), transforms.ToTensor(),
+ transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
+
+ if args.dataset == 'Pose_300W_LP':
+ pose_dataset = datasets.Pose_300W_LP(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFLW2000':
+ pose_dataset = datasets.AFLW2000(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'BIWI':
+ pose_dataset = datasets.BIWI(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFLW':
+ pose_dataset = datasets.AFLW(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFLW_aug':
+ pose_dataset = datasets.AFLW_aug(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFW':
+ pose_dataset = datasets.AFW(args.data_dir, args.filename_list, transformations)
+ else:
+ print 'Error: not a valid dataset name'
+ sys.exit()
+ train_loader = torch.utils.data.DataLoader(dataset=pose_dataset,
+ batch_size=batch_size,
+ shuffle=True,
+ num_workers=2)
+
+ model.cuda(gpu)
+ softmax = nn.Softmax()
+ criterion = nn.CrossEntropyLoss().cuda()
+ reg_criterion = nn.MSELoss().cuda()
+ smooth_l1_loss = nn.SmoothL1Loss().cuda()
+ # Regression loss coefficient
+ alpha = args.alpha
+
+ optimizer = torch.optim.Adam([{'params': get_ignored_params(model), 'lr': 0},
+ {'params': get_non_ignored_params(model), 'lr': args.lr},
+ {'params': get_fc_params(model), 'lr': args.lr}],
+ lr = args.lr)
+
+ print 'Ready to train network.'
+
+ print 'Second phase of training (finetuning layer).'
+ for epoch in range(num_epochs_ft):
+ for i, (images, labels, cont_labels, name) in enumerate(train_loader):
+ images = Variable(images.cuda(gpu))
+
+ label_angles = Variable(cont_labels[:,:3].cuda(gpu))
+
+ optimizer.zero_grad()
+ model.zero_grad()
+
+ pre_yaw, pre_pitch, pre_roll, preangles, final_angles = model(images)
+
+ # Finetuning loss
+ loss_seq = []
+
+ loss_angles = smooth_l1_loss(final_angles, label_angles)
+ loss_seq.append(loss_angles)
+
+ grad_seq = [torch.Tensor(1).cuda(gpu) for _ in range(len(loss_seq))]
+ torch.autograd.backward(loss_seq, grad_seq)
+ optimizer.step()
+
+ if (i+1) % 100 == 0:
+ print ('Epoch [%d/%d], Iter [%d/%d] Losses: finetuning %.4f'
+ %(epoch+1, num_epochs_ft, i+1, len(pose_dataset)//batch_size, loss_angles.data[0]))
+
+ # Save models at numbered epochs.
+ if epoch % 1 == 0 and epoch < num_epochs_ft:
+ print 'Taking snapshot...'
+ torch.save(model.state_dict(),
+ 'output/snapshots/' + args.output_string + '_epoch_'+ str(epoch+1) + '.pkl')
diff --git a/code/train_hopenet_new.py b/code/train_hopenet_new.py
new file mode 100644
index 0000000..988f58f
--- /dev/null
+++ b/code/train_hopenet_new.py
@@ -0,0 +1,221 @@
+import numpy as np
+import torch
+import torch.nn as nn
+from torch.autograd import Variable
+from torch.utils.data import DataLoader
+from torchvision import transforms
+import torchvision
+import torch.backends.cudnn as cudnn
+import torch.nn.functional as F
+
+import cv2
+import matplotlib.pyplot as plt
+import sys
+import os
+import argparse
+
+import datasets
+import hopenet
+import torch.utils.model_zoo as model_zoo
+
+model_urls = {
+ 'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',
+ 'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',
+ 'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',
+ 'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',
+ 'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',
+}
+
+def parse_args():
+ """Parse input arguments."""
+ parser = argparse.ArgumentParser(description='Head pose estimation using the Hopenet network.')
+ parser.add_argument('--gpu', dest='gpu_id', help='GPU device id to use [0]',
+ default=0, type=int)
+ parser.add_argument('--num_epochs', dest='num_epochs', help='Maximum number of training epochs.',
+ default=5, type=int)
+ parser.add_argument('--num_epochs_ft', dest='num_epochs_ft', help='Maximum number of finetuning epochs.',
+ default=5, type=int)
+ parser.add_argument('--batch_size', dest='batch_size', help='Batch size.',
+ default=16, type=int)
+ parser.add_argument('--lr', dest='lr', help='Base learning rate.',
+ default=0.001, type=float)
+ parser.add_argument('--data_dir', dest='data_dir', help='Directory path for data.',
+ default='', type=str)
+ parser.add_argument('--filename_list', dest='filename_list', help='Path to text file containing relative paths for every example.',
+ default='', type=str)
+ parser.add_argument('--output_string', dest='output_string', help='String appended to output snapshots.', default = '', type=str)
+ parser.add_argument('--alpha', dest='alpha', help='Regression loss coefficient.',
+ default=0.001, type=float)
+ parser.add_argument('--iter_ref', dest='iter_ref', help='Number of iterative refinement passes.',
+ default=1, type=int)
+ parser.add_argument('--dataset', dest='dataset', help='Dataset type.', default='Pose_300W_LP', type=str)
+ args = parser.parse_args()
+ return args
+
+def get_ignored_params(model):
+ # Generator function that yields ignored params.
+ b = []
+ b.append(model.conv1)
+ b.append(model.bn1)
+ for i in range(len(b)):
+ for module_name, module in b[i].named_modules():
+ if 'bn' in module_name:
+ module.eval()
+ for name, param in module.named_parameters():
+ yield param
+
+def get_non_ignored_params(model):
+ # Generator function that yields params that will be optimized.
+ b = []
+ b.append(model.layer1)
+ b.append(model.layer2)
+ b.append(model.layer3)
+ b.append(model.layer4)
+
+ for i in range(len(b)):
+ for module_name, module in b[i].named_modules():
+ if 'bn' in module_name:
+ module.eval()
+ for name, param in module.named_parameters():
+ yield param
+
+def get_fc_params(model):
+ b = []
+ b.append(model.fc_yaw)
+ b.append(model.fc_pitch)
+ b.append(model.fc_roll)
+ b.append(model.lstm)
+ b.append(model.fc_lstm)
+ for i in range(len(b)):
+ for module_name, module in b[i].named_modules():
+ for name, param in module.named_parameters():
+ yield param
+
+def load_filtered_state_dict(model, snapshot):
+ # By user apaszke from discuss.pytorch.org
+ model_dict = model.state_dict()
+ # 1. filter out unnecessary keys
+ snapshot = {k: v for k, v in snapshot.items() if k in model_dict}
+ # 2. overwrite entries in the existing state dict
+ model_dict.update(snapshot)
+ # 3. load the new state dict
+ model.load_state_dict(model_dict)
+
+if __name__ == '__main__':
+ args = parse_args()
+
+ cudnn.enabled = True
+ num_epochs = args.num_epochs
+ num_epochs_ft = args.num_epochs_ft
+ batch_size = args.batch_size
+ gpu = args.gpu_id
+
+ if not os.path.exists('output/snapshots'):
+ os.makedirs('output/snapshots')
+
+ model = hopenet.Hopenet_new(torchvision.models.resnet.Bottleneck, [3, 4, 6, 3], 66)
+
+ load_filtered_state_dict(model, model_zoo.load_url(model_urls['resnet50']))
+
+ print 'Loading data.'
+
+ transformations = transforms.Compose([transforms.Scale(240),
+ transforms.RandomCrop(224), transforms.ToTensor(),
+ transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
+
+ if args.dataset == 'Pose_300W_LP':
+ pose_dataset = datasets.Pose_300W_LP(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFLW2000':
+ pose_dataset = datasets.AFLW2000(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'BIWI':
+ pose_dataset = datasets.BIWI(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFLW':
+ pose_dataset = datasets.AFLW(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFLW_aug':
+ pose_dataset = datasets.AFLW_aug(args.data_dir, args.filename_list, transformations)
+ elif args.dataset == 'AFW':
+ pose_dataset = datasets.AFW(args.data_dir, args.filename_list, transformations)
+ else:
+ print 'Error: not a valid dataset name'
+ sys.exit()
+ train_loader = torch.utils.data.DataLoader(dataset=pose_dataset,
+ batch_size=batch_size,
+ shuffle=True,
+ num_workers=2)
+
+ model.cuda(gpu)
+ softmax = nn.Softmax()
+ criterion = nn.CrossEntropyLoss().cuda()
+ reg_criterion = nn.MSELoss().cuda()
+ smooth_l1_loss = nn.SmoothL1Loss().cuda()
+ # Regression loss coefficient
+ alpha = args.alpha
+
+ idx_tensor = [idx for idx in xrange(66)]
+ idx_tensor = Variable(torch.FloatTensor(idx_tensor)).cuda(gpu)
+
+ optimizer = torch.optim.Adam([{'params': get_ignored_params(model), 'lr': 0},
+ {'params': get_non_ignored_params(model), 'lr': args.lr},
+ {'params': get_fc_params(model), 'lr': args.lr * 5}],
+ lr = args.lr)
+
+ print 'Ready to train network.'
+
+ print 'Second phase of training (finetuning layer).'
+ for epoch in range(num_epochs_ft):
+ for i, (images, labels, cont_labels, name) in enumerate(train_loader):
+ images = Variable(images.cuda(gpu))
+ label_yaw = Variable(labels[:,0].cuda(gpu))
+ label_pitch = Variable(labels[:,1].cuda(gpu))
+ label_roll = Variable(labels[:,2].cuda(gpu))
+
+ label_angles = Variable(cont_labels[:,:3].cuda(gpu))
+ label_yaw_cont = Variable(cont_labels[:,0].cuda(gpu))
+ label_pitch_cont = Variable(cont_labels[:,1].cuda(gpu))
+ label_roll_cont = Variable(cont_labels[:,2].cuda(gpu))
+
+ optimizer.zero_grad()
+ model.zero_grad()
+
+ pre_yaw, pre_pitch, pre_roll, preangles, final_angles = model(images)
+
+ # Cross entropy loss
+ loss_yaw = criterion(pre_yaw, label_yaw)
+ loss_pitch = criterion(pre_pitch, label_pitch)
+ loss_roll = criterion(pre_roll, label_roll)
+
+ # MSE loss
+ yaw_predicted = preangles[0]
+ pitch_predicted = preangles[1]
+ roll_predicted = preangles[2]
+
+ loss_reg_yaw = reg_criterion(yaw_predicted, label_yaw_cont)
+ loss_reg_pitch = reg_criterion(pitch_predicted, label_pitch_cont)
+ loss_reg_roll = reg_criterion(roll_predicted, label_roll_cont)
+
+ # Total loss
+ loss_yaw += alpha * loss_reg_yaw
+ loss_pitch += alpha * loss_reg_pitch
+ loss_roll += alpha * loss_reg_roll
+
+ # LSTM loss
+ loss_seq = [loss_yaw, loss_pitch, loss_rol]
+ loss_lstm = reg_criterion(final_angles, label_angles)
+ loss_seq.append(loss_lstm)
+
+ grad_seq = [torch.Tensor(1).cuda(gpu) for _ in range(len(loss_seq))]
+ torch.autograd.backward(loss_seq, grad_seq)
+ optimizer.step()
+
+ if (i+1) % 100 == 0:
+ print ('Epoch [%d/%d], Iter [%d/%d] Losses: pre-yaw %.4f, pre-pitch %.4f, pre-roll %.4f, finetuning %.4f'
+ %(epoch+1, num_epochs_ft, i+1, len(pose_dataset)//batch_size, loss_yaw.data[0], loss_pitch.data[0], loss_roll.data[0], loss_angles.data[0]))
+ # if epoch == 0:
+ # torch.save(model.state_dict(),
+ # 'output/snapshots/' + args.output_string + '_iter_'+ str(i+1) + '.pkl')
+
+ # Save models at numbered epochs.
+ if epoch % 1 == 0 and epoch < num_epochs_ft:
+ print 'Taking snapshot...'
+ torch.save(model.state_dict(),
+ 'output/snapshots/' + args.output_string + '_epoch_'+ str(num_epochs+epoch+1) + '.pkl')
--
Gitblit v1.8.0