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