def save_prediction(self, name, perceptual_loss_net):
super(vae_wrapper, self).save_prediction(name, perceptual_loss_net)
eval_prep = self.eval_prep
eval_input = self.train_prep.scale_input(
eval_prep.inputs) # .scale_input(eval_prep.inputs)
eval_output = self.train_prep.scale_output(
eval_prep.outputs) # scale_output(eval_prep.outputs)
local_positions = to_numpy(self.final_outputs)
bone_dependencies, global_positions, rotations = eval_prep.get_global_pos_from_prediction(
eval_input, local_positions, self.train_prep)
_, reference_positions, rotations = eval_prep.get_global_pos_from_prediction(
eval_input, eval_output, self.train_prep)
self.export_anim_unity(
self.train_prep.scale_back_output(local_positions), name)
self.stats_printer(self.train_prep.scale_back_output(local_positions),
eval_prep.outputs, name, perceptual_loss_net)
self.save_anim(global_positions, reference_positions,
bone_dependencies, rotations, name)
def save_prediction(self, name, perceptual_loss_net):
super(rnn_wrapper_2, self).save_prediction(name, perceptual_loss_net)
training_prep = self.train_prep
eval_prep = self.eval_prep
final_outputs = self.final_outputs
eval_input = training_prep.scale_input(
eval_prep.inputs) # .scale_input(eval_prep.inputs)
eval_output = training_prep.scale_output(
eval_prep.outputs) # scale_output(eval_prep.outputs)
idx1 = (eval_input.shape[0]) % eval_prep.total_seq_length
idx2 = idx1 + final_outputs.shape[0]
eval_input = eval_input[idx1:idx2, :]
eval_output = eval_output[idx1:idx2, :]
self.final_outputs = self.final_outputs.squeeze(1)
local_positions = to_numpy(self.final_outputs)
bone_dependencies, global_positions, rotations = eval_prep.get_global_pos_from_prediction(
eval_input,
local_positions,
self.train_prep,
start_idx=idx1,
end_idx=idx2)
_, reference_positions, rotations = eval_prep.get_global_pos_from_prediction(
eval_input,
eval_output,
self.train_prep,
start_idx=idx1,
end_idx=idx2)
self.export_anim_unity(
self.train_prep.scale_back_output(local_positions), name)
self.stats_printer(self.train_prep.scale_back_output(local_positions),
eval_prep.outputs, name, perceptual_loss_net)
self.save_anim(global_positions, reference_positions,
bone_dependencies, rotations, name)
x_len = eval_x.shape[1]
final_outputs = glow_model.predict(eval_cond[:, :, :500], STACKCOUNT, x_len)
#
eval_input = training_prep.scale_input(
eval_prep.inputs) # .scale_input(eval_prep.inputs)
eval_output = training_prep.scale_output(
eval_prep.outputs) # scale_output(eval_prep.outputs)
idx1 = (eval_input.shape[0]) % eval_prep.total_seq_length
idx2 = idx1 + 499
eval_input = eval_input[idx1:idx2, :]
eval_output = eval_output[idx1:idx2, :]
bone_dependencies, global_positions, rotations = eval_prep.get_global_pos_from_prediction(
eval_input,
to_numpy(final_outputs),
training_prep,
start_idx=idx1,
end_idx=idx1 + 499)
bone_dependencies, reference_positions, rotations = eval_prep.get_global_pos_from_prediction(
eval_input, eval_output, training_prep, start_idx=idx1, end_idx=idx2)
MOTIONTYPE = "Boxing"
# sp.print_stats(global_positions, reference_positions, ["l_hand", "r_hand", "l_shoulder", "r_shoulder", "hip", "l_foot", "r_foot", "l_elbow", "r_elbow", "l_knee", "r_knee"], MOTIONTYPE)
if __name__ == '__main__':
anim = Animator.MocapAnimator(global_positions, [''] * 40,
bone_dependencies,
1.0 / TARGET_FPS,
heading_dirs=rotations,
name="trained.avi")
rnnvae_model.train_model(input=train_feet_lower_input,
conditional_input=train_feet_lower_conditional_input,
output=train_feet_output,
eval_input=eval_feet_lower_input,
eval_conditional_input=eval_feet_conditional_input,
eval_output=eval_feet_output,
learning_rate=learning_rate,
epochs=num_epochs,
batch_size=batch_size)
ff_outputs = ff_model.predict(eval_input)
final_outputs = rnnvae_model.predict(eval_feet_input, ff_outputs, STACKCOUNT)
bone_dependencies, global_positions, _ = eval_prep.get_global_pos_from_prediction(
eval_input, to_numpy(final_outputs), training_prep)
_, reference_positions, _ = eval_prep.get_global_pos_from_prediction(
eval_input, eval_output, training_prep)
MOTIONTYPE = "Boxing"
sp.print_stats(global_positions, reference_positions, [
"l_hand", "r_hand", "l_shoulder", "r_shoulder", "hip", "l_foot", "r_foot",
"l_elbow", "r_elbow", "l_knee", "r_knee"
], MOTIONTYPE)
if __name__ == '__main__':
anim = Animator.MocapAnimator(
global_positions, [''] * 40,
bone_dependencies,
1.0 / TARGET_FPS,
heading_dirs=eval_prep.heading_dirs[-global_positions.shape[0]:],