retargeting on win

retargeting/demo-win.py

@@ -0,0 +1,48 @@
+import os
+from datasets.bvh_parser import BVH_file
+from datasets.bvh_writer import BVH_writer
+from models.IK import fix_foot_contact
+from os.path import join as pjoin
+
+
+# downsampling and remove redundant joints
+def copy_ref_file(src, dst):
+    file = BVH_file(src)
+    writer = BVH_writer(file.edges, file.names)
+    writer.write_raw(file.to_tensor(quater=True)[..., ::2], 'quaternion', dst)
+
+
+def get_height(file):
+    file = BVH_file(file)
+    return file.get_height()
+
+
+def example(src_name, dest_name, bvh_name, test_type, output_path):
+    if not os.path.exists(output_path):
+        os.makedirs(output_path)
+
+    input_file = os.getcwd()+'\\datasets\\Mixamo\\{}\\{}'.format(src_name, bvh_name)
+    ref_file = os.getcwd()+'\\datasets\\Mixamo\\{}\\{}'.format(dest_name, bvh_name)
+    copy_ref_file(input_file, pjoin(output_path, 'input.bvh'))
+    copy_ref_file(ref_file, pjoin(output_path, 'gt.bvh'))
+    height = get_height(input_file)
+
+    bvh_name = bvh_name.replace(' ', '_')
+    input_file = os.getcwd()+'\\datasets\\Mixamo\\{}\\{}'.format(src_name, bvh_name)
+    ref_file = os.getcwd()+'\\datasets\\Mixamo\\{}\\{}'.format(dest_name, bvh_name)
+
+    cmd = 'python eval_single_pair.py --input_bvh={} --target_bvh={} --output_filename={} --test_type={}'.format(
+        input_file, ref_file, pjoin(output_path, 'result.bvh'), test_type
+    )
+    os.system(cmd)
+
+    fix_foot_contact(pjoin(output_path, 'result.bvh'),
+                     pjoin(output_path, 'input.bvh'),
+                     pjoin(output_path, 'result.bvh'),
+                     height)
+
+
+if __name__ == '__main__':
+    example('Aj', 'BigVegas', 'Dancing Running Man.bvh', 'intra', os.getcwd()+'\\examples\\intra_structure')
+    example('BigVegas', 'Mousey_m', 'Dual Weapon Combo.bvh', 'cross', os.getcwd()+'\\examples\\cross_structure')
+    print('Finished!')

retargeting/demo.sh

@@ -1 +1 @@
-python demo.py 
+python demo-win.py 

retargeting/eval_single_pair.py

@@ -1,16 +1,20 @@
 import os
+import platform
 import torch
 from models import create_model
 from datasets import create_dataset
 import option_parser
 
+separator = '/'
+if platform.system() == 'Windows':
+    separator = '\\'
 
 def eval_prepare(args):
     character = []
     file_id = []
     character_names = []
-    character_names.append(args.input_bvh.split('/')[-2])
-    character_names.append(args.target_bvh.split('/')[-2])
+    character_names.append(args.input_bvh.split(separator)[-2])
+    character_names.append(args.target_bvh.split(separator)[-2])
     if args.test_type == 'intra':
         if character_names[0].endswith('_m'):
             character = [['BigVegas', 'BigVegas'], character_names]
@@ -35,9 +39,9 @@ def eval_prepare(args):
 
 
 def recover_space(file):
-    l = file.split('/')
+    l = file.split(separator)
     l[-1] = l[-1].replace('_', ' ')
-    return '/'.join(l)
+    return separator.join(l)
 
 
 def main():
@@ -54,9 +58,16 @@ def main():
     args.target_bvh = recover_space(args.target_bvh)
     args.output_filename = recover_space(args.output_filename)
 
+    # print(input_bvh)
+    # print(input_bvh)
+    # print(input_bvh)
+    # print(target_bvh)
+    # print(target_bvh)
+    # print(target_bvh)
+
     character_names, file_id, src_id = eval_prepare(args)
-    input_character_name = args.input_bvh.split('/')[-2]
-    output_character_name = args.target_bvh.split('/')[-2]
+    input_character_name = args.input_bvh.split(separator)[-2]
+    output_character_name = args.target_bvh.split(separator)[-2]
     output_filename = args.output_filename
 
     test_device = args.cuda_device
@@ -91,7 +102,10 @@ def main():
     model.set_input(input_motion)
     model.test()
 
-    os.system('cp "{}/{}/0_{}.bvh" "./{}"'.format(model.bvh_path, output_character_name, src_id, output_filename))
+    if platform.system() == 'Windows':
+        os.system('cp "{}\\{}\\0_{}.bvh" "{}"'.format(model.bvh_path, output_character_name, src_id, output_filename))
+    else:
+        os.system('cp "{}/{}/0_{}.bvh" "./{}"'.format(model.bvh_path, output_character_name, src_id, output_filename))
 
 
 if __name__ == '__main__':

retargeting/models/architecture.py

@@ -6,7 +6,11 @@
 from option_parser import try_mkdir
 
 import os
+import platform
 
+separator = '/'
+if platform.system() == 'Windows':
+    separator = '\\'
 
 class GAN_model(BaseModel):
     def __init__(self, args, character_names, dataset):
@@ -42,7 +46,7 @@ def __init__(self, args, character_names, dataset):
             for i in range(self.n_topology):
                 self.err_crit.append(Eval_Criterion(dataset.joint_topologies[i]))
             self.id_test = 0
-            self.bvh_path = os.path.join(args.save_dir, 'results/bvh')
+            self.bvh_path = os.path.join(args.save_dir, 'results'+separator+'bvh')
             option_parser.try_mkdir(self.bvh_path)
 
             self.writer = []

retargeting/option_parser.py

@@ -1,9 +1,15 @@
 import argparse
 
+import os
+import platform
+
+separator = '/'
+if platform.system() == 'Windows':
+    separator = '\\'
 
 def get_parser():
     parser = argparse.ArgumentParser()
-    parser.add_argument('--save_dir', type=str, default='./pretrained', help='directory for all savings')
+    parser.add_argument('--save_dir', type=str, default=os.getcwd()+separator+'pretrained', help='directory for all savings')
     parser.add_argument('--cuda_device', type=str, default='cuda:0', help='cuda device number, eg:[cuda:0]')
     parser.add_argument('--num_layers', type=int, default=2, help='number of layers')
     parser.add_argument('--learning_rate', type=float, default=2e-4, help='learning rate')
@@ -69,4 +75,7 @@ def get_std_bvh(args=None, dataset=None):
 def try_mkdir(path):
     import os
     if not os.path.exists(path):
-        os.system('mkdir -p {}'.format(path))
+        if platform.system() == 'Windows':
+            os.system('mkdir {}'.format(path))
+        else:
+            os.system('mkdir -p {}'.format(path))

retargeting/pretrained/para.txt

@@ -1 +1 @@
-train_new.py --save_dir=./pretrained/ --num_layers=2 --rotation=quaterneion --downsampling=stride2 --batch_normalization=0 --upsampling=linear --activation=LeakyReLU --kernel_size=15 --base_channel_num=-1 --normalization=1 --skeleton_dist=2 --verbose=1 --learning_rate=2e-4 --skeleton_pool=mean --data_augment=1 --padding_mode=reflection --cuda_device=cuda:0 --batch_size=256 --debug=0 --skeleton_info=concat --patch_gan=1 --fk_world=0 --eval=0 --ee_loss=vec --ee_loss_fact=learn --pos_repr=3d --model=mul_top_mul_ske --ee_velo=1 --rec_loss_mode=extra_global_pos --simple_operator=0 --epoch_num=30000 --gan_mode=lsgan --lambda_ee=100
+train_new.py --save_dir=D:\deeplearning\research_work\eb7\deep-motion-editing\retargeting\pretrained\ --num_layers=2 --rotation=quaterneion --downsampling=stride2 --batch_normalization=0 --upsampling=linear --activation=LeakyReLU --kernel_size=15 --base_channel_num=-1 --normalization=1 --skeleton_dist=2 --verbose=1 --learning_rate=2e-4 --skeleton_pool=mean --data_augment=1 --padding_mode=reflection --cuda_device=cuda:0 --batch_size=256 --debug=0 --skeleton_info=concat --patch_gan=1 --fk_world=0 --eval=0 --ee_loss=vec --ee_loss_fact=learn --pos_repr=3d --model=mul_top_mul_ske --ee_velo=1 --rec_loss_mode=extra_global_pos --simple_operator=0 --epoch_num=30000 --gan_mode=lsgan --lambda_ee=100

retargeting/test.py

@@ -1,11 +1,15 @@
 import os
+import platform
 from os.path import join as pjoin
 from get_error import full_batch
 import numpy as np
 from option_parser import try_mkdir
 from eval import eval
 import argparse
 
+separator = '/'
+if platform.system() == 'Windows':
+    separator = '\\'
 
 def batch_copy(source_path, suffix, dest_path, dest_suffix=None):
     try_mkdir(dest_path)
@@ -24,13 +28,13 @@ def batch_copy(source_path, suffix, dest_path, dest_suffix=None):
     test_characters = ['Mousey_m', 'Goblin_m', 'Mremireh_m', 'Vampire_m']
 
     parser = argparse.ArgumentParser()
-    parser.add_argument('--save_dir', type=str, default='./pretrained/')
+    parser.add_argument('--save_dir', type=str, default='.'+separator+'pretrained'+separator)
     args = parser.parse_args()
     prefix = args.save_dir
 
-    cross_dest_path = pjoin(prefix, 'results/cross_structure/')
-    intra_dest_path = pjoin(prefix, 'results/intra_structure/')
-    source_path = pjoin(prefix, 'results/bvh/')
+    cross_dest_path = pjoin(prefix, 'results'+separator+'cross_structure'+separator+'')
+    intra_dest_path = pjoin(prefix, 'results'+separator+'intra_structure'+separator+'')
+    source_path = pjoin(prefix, 'results'+separator+'bvh'+separator+'')
 
     cross_error = []
     intra_error = []
@@ -59,7 +63,7 @@ def batch_copy(source_path, suffix, dest_path, dest_suffix=None):
     cross_error_mean = cross_error.mean()
     intra_error_mean = intra_error.mean()
 
-    os.system('rm -r %s' % pjoin(prefix, 'results/bvh'))
+    os.system('rm -r %s' % pjoin(prefix, 'results'+separator+'bvh'))
 
     print('Intra-retargeting error:', intra_error_mean)
     print('Cross-retargeting error:', cross_error_mean)