diff --git a/cvae.py b/cvae.py index 9dfeb02..5c47d7e 100644 --- a/cvae.py +++ b/cvae.py @@ -1,5 +1,5 @@ import os -os.environ["CUDA_VISIBLE_DEVICES"] = "0" +os.environ["CUDA_VISIBLE_DEVICES"] = "2" import tensorflow as tf gpus = tf.config.experimental.list_physical_devices('GPU') @@ -21,7 +21,6 @@ import random import numpy as np import scipy as sp from custom_utils.custom_utils import UV2Mesh, maskFacesBatch, _maskFaces -from tensorboardX import SummaryWriter import plotly.graph_objects as go import plotly.express as ex from PIL import Image @@ -38,7 +37,7 @@ from custom_utils.smpl import SMPL from models.cvae_model import CVAE -name='cvae_16' +name='cvae_17' def init_wandb(): run = wandb.init(project='vae', name=name, notes='learn offsets. smaller flattened latent vectors. with 25-85k pretraining. smaller batchsize. fixed balancing of reconst and kl loss for baseline. \ @@ -46,7 +45,7 @@ new implementation for kl loss. no lr schedule. 25 jumpsuits, average reconst_lo offsets over average unposed garment. remove global orientation (in posed body input and in pose_shape parameters. \ retrained body encoder with deoriented bodies \ use mean l2 loss. small batch. l2 * 1000, balancing to 1 from 0.01\ -fixed gender issue and retrained body encoder\ +fixed gender issue and retrained body encoder, FINAL\ remove shape from unposed garment\ training on 100 garments with offsets over the average of these') artifact = wandb.Artifact('code', type='code') @@ -129,35 +128,29 @@ def train_step(batch): with tf.GradientTape() as tape: uv_curvature = tf.one_hot(batch['uv_curvature'], 7) - pred, mean, logvar, z = model([batch['uv_unposed'], batch['uv_static'], batch['uv_body_posed'], uv_curvature, batch['uv_mask'], batch['uv_shaped_body']]) + pred, mean, logvar, z = model([batch['uv_unposed'], batch['uv_static'], batch['uv_body_posed'], uv_curvature, batch['uv_mask']], training=True, use_mask=True) - mesh_pred = uv2mesh(tf.transpose(pred, [0,3,1,2])) - mesh_pred_posed = smpl.run_garment(batch['pose_shape'][:, :72], batch['pose_shape'][:, 72:], mesh_pred) + mesh_pred = uv2mesh(pred) mesh_orig = batch['mesh_unposed'] shape = batch['pose_shape'][:, 72:] gender = tf.cast(tf.greater_equal(shape[:, 10] , 0.5), tf.int32) shape_offsets = tf.stack([tf.gather(tf.tensordot(shape[i,:10][tf.newaxis, :], smpl.shapedirs[gender[i]], axes=[[1], [2]])[0,:,:], smpl.to_keep, axis=0) for i in range(shape.shape[0])]) - mesh_pred_posed = mesh_pred_posed + shape_offsets * batch['mesh_mask'][:,:,tf.newaxis] + mesh_pred_shaped = mesh_pred + shape_offsets * batch['mesh_mask'][:,:,tf.newaxis] + mesh_pred_posed = smpl.run_garment(batch['pose_shape'][:, :72], batch['pose_shape'][:, 72:], mesh_pred_shaped) e_loss = tf.reduce_mean(edge_loss(mesh_pred, mesh_orig, batch['mesh_mask'])) c_loss = tf.reduce_mean(collision_penalty(mesh_pred_posed, batch['mesh_body_posed'], batch['mesh_mask'])) l1_loss = tf.reduce_sum(tf.reduce_sum((tf.math.abs(tf.transpose(batch['uv_unposed'], [0,2,3,1]) - pred)), axis=-1) * batch['uv_mask'], axis=[1, 2]) l2_loss = 1e3 * tf.reduce_mean(tf.reduce_sum(((tf.transpose(batch['uv_unposed'], [0,2,3,1]) - pred) ** 2), axis=-1) * batch['uv_mask'], axis=[1, 2]) # ms_ssim = tf.image.ssim_multiscale(tf.transpose(pred, [0,2,3,1]) + 1, tf.transpose(batch['uv_unposed'], [0,2,3,1]) + 1, max_val=2) - ms_ssim = tf.image.ssim_multiscale(pred + 1, tf.transpose(batch['uv_unposed'], [0,2,3,1]) + 1, max_val=3) + ms_ssim = tf.image.ssim_multiscale(pred + 1, tf.transpose(batch['uv_unposed'], [0,2,3,1]) + 1, max_val=3, power_factors=(0.5, 0.2, 0.1, 0.1, 0.1)) # ms_ssim = tf.reduce_mean(tf.image.ssim_multiscale((tf.transpose(pred, [0,2,3,1]) + 1) * batch['uv_mask'][:,:,:,tf.newaxis], (tf.transpose(batch['uv_unposed'], [0,2,3,1]) + 1) * batch['uv_mask'][:,:,:,tf.newaxis], max_val=2)) - rfft_loss = 0 #tf.reduce_mean(tf.reduce_sum(tf.math.abs(tf.signal.rfft3d(pred) - tf.signal.rfft3d(batch['uv_unposed'])), axis=[1,2,3])) - fft_loss = tf.reduce_sum(tf.math.abs(tf.signal.fft3d(tf.complex(pred, tf.zeros_like(pred))) - tf.signal.fft3d(tf.complex(tf.transpose(batch['uv_unposed'], [0,2,3,1]), tf.zeros_like(tf.transpose(batch['uv_unposed'], [0,2,3,1]))))), axis=[1,2,3]) - fft_loss_pc, fft_loss_pc_low, fft_loss_pc_high = 0,0,0 #fft_loss_per_channel(pred, batch['uv_unposed']) + fft_loss = 0 #tf.reduce_sum(tf.math.abs(tf.signal.fft3d(tf.complex(pred, tf.zeros_like(pred))) - tf.signal.fft3d(tf.complex(tf.transpose(batch['uv_unposed'], [0,2,3,1]), tf.zeros_like(tf.transpose(batch['uv_unposed'], [0,2,3,1]))))), axis=[1,2,3]) tv_loss = tf.image.total_variation(pred) reconstr_loss = l2_loss reconstr_loss = tf.reduce_mean(reconstr_loss) - # logpx_z = -reconstr_loss - # logpz = log_normal_pdf(z[:,:model.latent_dim], 0., 0.) - # logqz_x = log_normal_pdf(z[:,:model.latent_dim], mean, logvar) - # loss = -tf.reduce_mean(logpx_z + balancing_coeff * (logpz - logqz_x)) kl = kl_loss(mean, logvar) loss = reconstr_loss + balancing_coeff * kl # loss = reconstr_loss + balancing_coeff * kl @@ -178,8 +171,8 @@ def train_step(batch): # gradients = [tf.clip_by_value(g, -1.0, 1.0) for g in gradients] optimizer.apply_gradients(zip(gradients, model.trainable_variables)) return { 'train/loss': loss, 'train/edge_loss': e_loss, 'train/collision_loss': c_loss, 'train/l1_loss': l1_loss, 'train/euclid_error': metric, 'train/ms_ssim': ms_ssim, - 'train/fft_loss': fft_loss, 'train/euclid_error_reconst': metric_2, 'train/rfft_loss': rfft_loss, 'train/fft_oc_low': fft_loss_pc_low, 'train/fft_pc_high': fft_loss_pc_high, - 'train/fft_loss_per_channel': fft_loss_pc, 'train/euclid_error_posed': euclid_posed, 'train/reconstr_loss': reconstr_loss, 'train/tv_loss': tv_loss, + 'train/fft_loss': fft_loss, 'train/euclid_error_reconst': metric_2, + 'train/euclid_error_posed': euclid_posed, 'train/reconstr_loss': reconstr_loss, 'train/tv_loss': tv_loss, 'train/mean': mean, 'train/logvar': logvar, 'train/kl_loss': kl, 'train/reconst_loss_norm': reconstr_loss / (2 * (gamma ** 2)), 'train/var': np.e ** logvar, 'train/reg_coeff': reg_coeff, 'train/l2_loss': l2_loss } @@ -226,9 +219,9 @@ def valid_step(batch): def plot_everything(batch: dict, type: str): uv_curvature = tf.one_hot(batch['uv_curvature'], 7) - pred, mean, logvar, z = model([batch['uv_unposed'], batch['uv_static'], batch['uv_body_posed'], uv_curvature, batch['uv_mask'], batch['uv_shaped_body']]) + pred, mean, logvar, z = model([batch['uv_unposed'], batch['uv_static'], batch['uv_body_posed'], uv_curvature, batch['uv_mask']], training=False, use_mask=True) - mesh_pred = uv2mesh(tf.transpose(pred, [0,3,1,2])) + mesh_pred = uv2mesh(pred).numpy() mesh_orig = batch['mesh_unposed'] orig_img = tf.transpose(batch['uv_unposed'], [0,2,3,1]) @@ -237,20 +230,23 @@ def plot_everything(batch: dict, type: str): pred_img = pred img = tf.concat((uv_static_img, uv_body_posed_img, orig_img, pred_img), axis=2) imgs = [wandb.Image((img[i] + 1) / 2.0) for i in range(4)] - - F = np.load("custom_utils/smpl_faces.npy") - mask = batch['mesh_mask'].numpy().astype(bool) - garment_unposed, faces = maskFacesBatch(mesh_pred.numpy(), F, mask) - gt = [mesh_orig[i].numpy()[mask[i], :] for i in range(pred.shape[0])] - body = batch['mesh_body_posed'].numpy() - mesh_posed = smpl.run_garment(batch['pose_shape'][:, :72], batch['pose_shape'][:, 72:], mesh_pred).numpy() - shape = batch['pose_shape'][:, 72:] gender = tf.cast(tf.greater_equal(shape[:, 10] , 0.5), tf.int32) shape_offsets = tf.stack([tf.gather(tf.tensordot(shape[i,:10][tf.newaxis, :], smpl.shapedirs[gender[i]], axes=[[1], [2]])[0,:,:], smpl.to_keep, axis=0) for i in range(shape.shape[0])]) - garment_posed = [mesh_posed[i, mask[i], :] + shape_offsets.numpy()[i, mask[i], :] for i in range(body.shape[0])] + F = np.load("custom_utils/smpl_faces.npy") + mask = batch['mesh_mask'].numpy().astype(bool) + garment_unposed, faces = maskFacesBatch(mesh_pred, F, mask) + garment_unposed = [garment_unposed[i] + shape_offsets.numpy()[i, mask[i], :] for i in range(mask.shape[0])] + gt = [mesh_orig[i].numpy()[mask[i], :] + shape_offsets.numpy()[i, mask[i], :] for i in range(pred.shape[0])] + + body = batch['mesh_body_posed'].numpy() + for i in range(body.shape[0]): + mesh_pred[i, mask[i], :] += shape_offsets.numpy()[i, mask[i], :] + mesh_posed = smpl.run_garment(batch['pose_shape'][:, :72], batch['pose_shape'][:, 72:], mesh_pred).numpy() + + garment_posed = [mesh_posed[i, mask[i], :] for i in range(body.shape[0])] collision = [np.concatenate((body[i], garment_posed[i]), axis=0) for i in range(len(garment_posed))] collision_faces = [np.concatenate((F, faces[i] + 14475), axis=0) for i in range(len(garment_posed))] collision_color = [np.concatenate((np.ones_like(body[i], dtype=np.int32)*100, np.ones_like(garment_posed[i], dtype=np.int32)*230), axis=0) for i in range(len(garment_posed))] @@ -274,7 +270,7 @@ def sample(folder): uv_curvature = tf.one_hot(batch['uv_curvature'], 7) nr_of_samples=2 - pred = model.sample(batch['uv_static'][0], batch['uv_body_posed'][0], uv_curvature[0], batch['uv_mask'][0], nr_of_samples=nr_of_samples) + pred = model.sample_single(batch['uv_static'][0], batch['uv_body_posed'][0], uv_curvature[0], batch['uv_mask'][0], nr_of_samples=nr_of_samples) pose, shape = batch['pose_shape'][0, :72], batch['pose_shape'][0, 72:] pose, shape = tf.tile(pose[tf.newaxis,:], [nr_of_samples,1]), tf.tile(shape[tf.newaxis,:], [nr_of_samples,1]) @@ -291,7 +287,7 @@ def sample(folder): _, faces = _maskFaces(np.ones((14475, 3)), F, mask) - pred_zero = model.sample(batch['uv_static'][0], batch['uv_body_posed'][0], uv_curvature[0], batch['uv_mask'][0], nr_of_samples=nr_of_samples, zero_latent=True) + pred_zero = model.sample_single(batch['uv_static'][0], batch['uv_body_posed'][0], uv_curvature[0], batch['uv_mask'][0], nr_of_samples=nr_of_samples, zero_latent=True) mesh_pred_zero = uv2mesh(tf.transpose(pred_zero, [0,3,1,2])) mesh_posed_zero = smpl.run_garment(pose, shape, mesh_pred_zero).numpy() @@ -412,18 +408,18 @@ if __name__ == '__main__': # 'mesh_static': (14475, 3,), 'mesh_mask': (14475,), 'uv_mask': (512, 256,), - 'uv_shaped_body': (3, 512, 256,), + # 'uv_shaped_body': (3, 512, 256,), } - dataset_train = factory.get_dataset(folder='/home/workspace/local-data/train_100', batch_size=batch_size, signature=signature_dynamic) - static_dataset_train = StaticDataset(folder='/home/workspace/local-data/train_100', signature=signature_static) + dataset_train = factory.get_dataset(folder='/home/workspace/local-data/train_100', batch_size=batch_size, signature=signature_dynamic, use_filled=False) + static_dataset_train = StaticDataset(folder='/home/workspace/local-data/train_100', signature=signature_static, use_filled=False) dataset_valid = factory.get_dataset(folder='/home/workspace/local-data/jumpsuit_20/valid', batch_size=batch_size, signature=signature_dynamic) static_dataset_valid = StaticDataset(folder='/home/workspace/local-data/jumpsuit_20/valid', signature=signature_static) model = CVAE(avg_garment_path='custom_data/train_100_no_shape_avg_uv.npy', latent_dim=128, static_model='checkpoints/ae_static_4', - body_model='checpoints/ae_body_7', + body_model='checpoints/ae_body_8', curvature_model='checpoints/ae_curvature_7') diff --git a/models/cvae_model.py b/models/cvae_model.py index efb9a3e..342f265 100644 --- a/models/cvae_model.py +++ b/models/cvae_model.py @@ -126,7 +126,7 @@ class CVAE(tf.keras.Model): # self.shuffle = tf.keras.layers.Lambda(lambda x: tf.stack([tf.gather(x[i], tf.random.shuffle(tf.range(x[i].shape[0]))) for i in range(x.shape[0])])) # @tf.function - def sample(self, uv_template, uv_body_posed, uv_curv, uv_mask, nr_of_samples=4, fix_noise=False, zero_latent=False): + def sample_single(self, uv_template, uv_body_posed, uv_curv, uv_mask, nr_of_samples=4, fix_noise=False): if not fix_noise: eps = tf.random.normal(shape=(nr_of_samples, self.latent_dim)) else: @@ -147,9 +147,6 @@ class CVAE(tf.keras.Model): z_cond = tf.concat([z_static, z_body, z_curv], axis=-1) z_cond = self.cond_encoder(z_cond) - if zero_latent: - eps = tf.zeros_like(eps) - z = tf.concat([eps, z_cond], axis=-1) x = self.decoder(z) @@ -159,6 +156,26 @@ class CVAE(tf.keras.Model): return x + def sample(self, uv_template, uv_body_posed, uv_curv, uv_mask=None, noise=None): + if noise is None: + eps = tf.random.normal(shape=(uv_template.shape[0], self.latent_dim)) + else: + eps = noise + + z_cond = self.encode_conds(uv_template, uv_body_posed, uv_curv) + z = tf.concat([eps, z_cond], axis=-1) + + x = self.decoder(z) + + if self.avg_garment is not None: + x = x + self.avg_garment[tf.newaxis,:,:,:] + + if uv_mask is not None: + x = x * uv_mask[:,:,:,tf.newaxis] + + return x + + def reparameterize(self, mean, logvar, fix_noise=False): if not fix_noise: eps = tf.random.normal(shape=mean.shape) @@ -176,22 +193,37 @@ class CVAE(tf.keras.Model): return logits - def call(self, inputs, training=None, zero_latent=False, fix_noise=False): - uv_unposed, uv_template, uv_body_posed, uv_curv = inputs - uv_unposed, uv_template, uv_body_posed = tf.transpose(uv_unposed, [0,2,3,1]), tf.transpose(uv_template, [0,2,3,1]), tf.transpose(uv_body_posed, [0,2,3,1]) + def encode_conds(self, uv_template, uv_body_posed, uv_curv): + if uv_template.shape[1] == 3: + uv_template = tf.transpose(uv_template, [0,2,3,1]) + if uv_body_posed.shape[1] == 3: + uv_body_posed = tf.transpose(uv_body_posed, [0,2,3,1]) + if uv_curv.shape[1] == 3: + uv_curv = tf.transpose(uv_curv, [0,2,3,1]) + z_static = self.static_enc(uv_template) z_body = self.body_enc(uv_body_posed) z_curv = self.curv_enc(uv_curv) - x = tf.concat([uv_unposed, uv_template, uv_body_posed, uv_curv], axis=-1) - - mean, logvar = self.encoder(x) - z = self.reparameterize(mean, logvar, fix_noise) z_cond = tf.concat([z_static, z_body, z_curv], axis=-1) z_cond = self.cond_encoder(z_cond) - if zero_latent: - z = tf.zeros_like(z) + return z_cond + + + def call(self, inputs, training=None, use_mask=False): + if use_mask: + uv_unposed, uv_template, uv_body_posed, uv_curv, uv_mask = inputs + else: + uv_unposed, uv_template, uv_body_posed, uv_curv = inputs + uv_unposed, uv_template, uv_body_posed = tf.transpose(uv_unposed, [0,2,3,1]), tf.transpose(uv_template, [0,2,3,1]), tf.transpose(uv_body_posed, [0,2,3,1]) + + x = tf.concat([uv_unposed, uv_template, uv_body_posed, uv_curv], axis=-1) + + mean, logvar = self.encoder(x) + z = self.reparameterize(mean, logvar) + + z_cond = self.encode_conds(uv_template, uv_body_posed, uv_curv) z = tf.concat([z, z_cond], axis=-1) @@ -200,14 +232,14 @@ class CVAE(tf.keras.Model): # z = tf.concat([z, z_curv], axis=-1) x = self.decoder(z) - # x = tf.transpose(x, [0,3,1,2]) # x = x * uv_mask[:,tf.newaxis,:,:] # x = tf.image.resize(x, (512, 256), tf.image.ResizeMethod.LANCZOS5) # x = x + tf.transpose(uv_shaped_body, [0,2,3,1]) if self.avg_garment is not None: x = x + self.avg_garment[tf.newaxis,:,:,:] - # x = x * uv_mask[:,:,:,tf.newaxis] + if use_mask: + x = x * uv_mask[:,:,:,tf.newaxis] return x, mean, logvar, z diff --git a/sample.py b/sample.py index ccac5d6..a09da1f 100644 --- a/sample.py +++ b/sample.py @@ -174,7 +174,7 @@ def sample(folder_main, folder_curv): uv_curvature = tf.math.argmax(pred, axis=-1) uv_curvature = tf.one_hot(uv_curvature, 7) - pred = model.sample(batch['uv_static'][0], + pred = model.sample_single(batch['uv_static'][0], batch['uv_body_posed'][0], uv_curvature, batch['uv_mask'][0], diff --git a/sample_pro.py b/sample_pro.py index b91f9ce..ed5d161 100644 --- a/sample_pro.py +++ b/sample_pro.py @@ -1,4 +1,3 @@ -from lib2to3.refactor import get_fixers_from_package import os os.environ["CUDA_VISIBLE_DEVICES"] = "0" import tensorflow as tf @@ -22,27 +21,12 @@ import random import numpy as np import scipy as sp from custom_utils.custom_utils import UV2Mesh, maskFacesBatch, _maskFaces -from tensorboardX import SummaryWriter import plotly.graph_objects as go import plotly.express as ex from PIL import Image +from plotly.subplots import make_subplots -from utils import str_to_bool -from tf_utils import check_tf_version, allow_memory_growth, split_gpu_for_testing -from load_models import load_generator, load_discriminator -from losses import d_logistic, d_logistic_r1_reg, g_logistic_non_saturating, g_logistic_ns_pathreg, edge_loss, uv_mask_loss, bend_loss -# from losses import collision_penalty as collision_penalty -from losses import collision_loss as collision_penalty -from convnext.models.convnext_tf import ConvNeXt, ConvNeXtReverse, Block -from stylegan2_ae.generator import Generator -from custom_utils.graph_utils import Reconstructor_V4_Decoder as GNNDecoder, sparse_to_tensor -from custom_utils.graph_utils import Reconstructor_v2 as SimpleReconstructor -from dataset import DatasetFactory, StaticDataset -from custom_utils.custom_utils import batch_to_dict from custom_utils.smpl import SMPL -from auto_encoder_body import AutoEncoder as BodyAE -from auto_encoder_curv import AutoEncoder as CurvatureAE -from auto_encoder_static import AutoEncoder as StaticAE from cvae_curv import CVAE as CVAECurv from models.cvae_model import CVAE as CVAEMain @@ -163,84 +147,251 @@ def get_frame(folder, nr): frames = sorted(os.listdir(pose_folder)) frames = [frame.split('.')[0] for frame in frames] # .split('_')[0] random.shuffle(frames) - return frames[0], frames[1:nr] + out = [frames[0]] + out.extend(frames[1:nr]) + return out -def get_bodies(folder, frame_main, frame_rand, nr=4, random=True): + +def get_bodies(folder, frames, random=True): bodies = [] - uv_posed_body_no_orient = np.load(os.path.join(folder, 'uv_posed_body_no_orient', frame_main + '.npz')); uv_posed_body_no_orient = uv_posed_body_no_orient['uv_posed_body_no_orient']; bodies.append(uv_posed_body_no_orient) - pass - -def get_curvs(folder, body, static, mask, frame=None, nr=4, random=True, model=None): - assert (random and model is not None) or (not random and model is None) - if random: - curv = model([body, static, mask]) - - - uv_curvature = tf.math.argmax(pred, axis=-1) - uv_curvature = tf.one_hot(uv_curvature, 7) - pass + poses_shapes = [] + uv_posed_body_no_orient = np.load(os.path.join(folder, 'uv_posed_body_no_orient', frames[0] + '.npz')); uv_posed_body_no_orient = uv_posed_body_no_orient['uv_posed_body_no_orient']; bodies.append(uv_posed_body_no_orient) + pose_shape = np.load(os.path.join(folder, 'pose_shape', frames[0] + '.npy')); poses_shapes.append(pose_shape) + if not random: + for _ in range(len(frames)-1): + bodies.append(uv_posed_body_no_orient) + poses_shapes.append(pose_shape) else: - curv = np.load(os.path.join(folder, 'uv_curvature')) #... - pass + for i in range(1, len(frames)): + uv_posed_body_no_orient = np.load(os.path.join(folder, 'uv_posed_body_no_orient', frames[i] + '.npz')); uv_posed_body_no_orient = uv_posed_body_no_orient['uv_posed_body_no_orient']; bodies.append(uv_posed_body_no_orient) + pose_shape = np.load(os.path.join(folder, 'pose_shape', frames[i] + '.npy')); poses_shapes.append(pose_shape) + bodies = np.stack(bodies) + poses_shapes = np.stack(poses_shapes).astype(np.float32) + return bodies, poses_shapes -def get_statics(folder, nr=4, random=True): - pass -def get_unposed(folder, nr=4, random=True): - pass +def get_curvs(folder, body, static, mask, frames, noise, random=True, model: CVAECurv=None, use_original=True): + assert (random and model is not None) or (not random) + uv_curvature_orig = np.load(os.path.join(folder, 'uv_curvature', frames[0] + '.npz')); uv_curvature_orig = uv_curvature_orig['curvature_uv'] -def plot_inputs(unposed, static, body, curv): - pass + noise = get_noise(not random, body.shape[0], latent_dim=128) + curv = model.sample(static, body, mask, noise) -def plot_variable_condition(condition, is_curv=False): - pass + uv_curvature = tf.math.argmax(curv, axis=-1).numpy() + if use_original: + uv_curvature[0,:,:] = uv_curvature_orig + out = tf.one_hot(uv_curvature, 7).numpy() -def plot_results(unposed, posed, body, mesh_mask): - body_color = np.ones_like(body, dtype=np.int32)*100 + return out +def get_statics(folder, frames, random=True): + statics = [] + uv_masks = [] + mesh_masks = [] + uv_static = np.load(os.path.join(folder, 'uv_static', frames[0].split('_')[0] + '.npz')); uv_static = uv_static['uv_static']; statics.append(uv_static) + uv_mask = np.load(os.path.join(folder, 'uv_mask', frames[0].split('_')[0] + '.npz')); uv_mask=uv_mask['uv_mask']; uv_mask = uv_mask.astype(np.float32); uv_masks.append(uv_mask) + mesh_mask = np.load(os.path.join(folder, 'mesh_mask', frames[0].split('_')[0] + '.npy')); mesh_masks.append(mesh_mask) + + if not random: + for _ in range(len(frames)-1): + statics.append(uv_static) + uv_masks.append(uv_mask) + mesh_masks.append(mesh_mask) + else: + for i in range(1, len(frames)): + uv_static = np.load(os.path.join(folder, 'uv_static', frames[i].split('_')[0] + '.npz')); uv_static = uv_static['uv_static']; statics.append(uv_static) + uv_mask = np.load(os.path.join(folder, 'uv_mask', frames[i].split('_')[0] + '.npz')); uv_mask=uv_mask['uv_mask']; uv_mask = uv_mask.astype(np.float32); uv_masks.append(uv_mask) + mesh_mask = np.load(os.path.join(folder, 'mesh_mask', frames[i].split('_')[0] + '.npy')); mesh_masks.append(mesh_mask) + + statics = np.stack(statics) + uv_masks = np.stack(uv_masks) + mesh_masks = np.stack(mesh_masks) + return statics, uv_masks, mesh_masks + + +def get_unposed(folder, frames, random=True): + uv_unposeds = [] + mesh_unposeds = [] + uv_unposed = np.load(os.path.join(folder, 'uv_posed', frames[0] + '.npz')); uv_unposed = uv_unposed['uv_gt']; uv_unposeds.append(uv_unposed) + mesh_unposed = np.load(os.path.join(folder, 'mesh_unposed_no_shape', frames[0] + '.npz')); mesh_unposed = mesh_unposed['mesh_unposed_no_shape']; mesh_unposeds.append(mesh_unposed) + if not random: + for _ in range(len(frames)-1): + uv_unposeds.append(uv_unposed) + mesh_unposeds.append(mesh_unposed) + else: + for i in range(1, len(frames)): + uv_static = np.load(os.path.join(folder, 'uv_static', frames[i].split('_')[0] + '.npz')); uv_static = uv_static['uv_static']; uv_unposeds.append(uv_static) + mesh_unposed = np.load(os.path.join(folder, 'mesh_unposed_no_shape', frames[i] + '.npz')); mesh_unposed = mesh_unposed['mesh_unposed_no_shape']; mesh_unposeds.append(mesh_unposed) + uv_unposeds = np.stack(uv_unposeds) + mesh_unposeds = np.stack(mesh_unposeds) + return uv_unposeds, mesh_unposeds + + +def make_square_2(figure): + figure.update_layout(scene_aspectmode='manual', scene_aspectratio=dict(x=1, y=1, z=1), + scene2_aspectmode='manual', scene2_aspectratio=dict(x=1, y=1, z=1)) + helper = dict(xaxis = dict(nticks=4, range=[-1,1],), + yaxis = dict(nticks=4, range=[-1,1],), + zaxis = dict(nticks=4, range=[-1,1],),) + figure.update_layout(scene = helper, + scene2 = helper) + return figure + +def make_square_4(figure): + figure.update_layout(scene_aspectmode='manual', scene_aspectratio=dict(x=1, y=1, z=1), + scene2_aspectmode='manual', scene2_aspectratio=dict(x=1, y=1, z=1), + scene3_aspectmode='manual', scene3_aspectratio=dict(x=1, y=1, z=1), + scene4_aspectmode='manual', scene4_aspectratio=dict(x=1, y=1, z=1)) + helper = dict(xaxis = dict(nticks=4, range=[-1,1],), + yaxis = dict(nticks=4, range=[-1,1],), + zaxis = dict(nticks=4, range=[-1,1],),) + figure.update_layout(scene = helper, + scene2 = helper, + scene3 = helper, + scene4 = helper,) + return figure + + +def plot_inputs(unposed, static, body, curv, mesh_mask): + curv = np.tile(curv.argmax(-1)[:,:,np.newaxis], [1,1,3])/6 + img = np.concatenate([unposed, static, body, curv], axis=-2) + ex.imshow((img + 1) / 2).show() + + mesh_static = uv2mesh(static[np.newaxis,:,:,:]).numpy()[0] + mesh_body = uv2mesh(body[np.newaxis,:,:,:]).numpy()[0] + fig = make_subplots(rows=1, cols=2, start_cell="top-left", specs = [ [{ 'type': 'scene' } for _ in range(2)] ]) F = np.load("custom_utils/smpl_faces.npy") _, faces = _maskFaces(np.ones((14475, 3)), F, mesh_mask) - pass + + x = mesh_static[mesh_mask, :] + y = mesh_body + fig.add_trace(go.Mesh3d( + x=x[:,0], + y=x[:,1], + z=x[:,2], + i=faces[:,0], + j=faces[:,1], + k=faces[:,2] + ), + row=1, col=1) + fig.add_trace(go.Mesh3d( + x=y[:,0], + y=y[:,2], + z=y[:,1], + i=F[:,0], + j=F[:,1], + k=F[:,2] + ), + row=1, col=2) + make_square_2(fig) + fig.show() + + +def plot_variable_condition(condition, mesh_mask, is_curv=False, is_body=False): + if is_curv: + img = np.concatenate(condition.argmax(-1), axis=-1) + ex.imshow(img / img.max()).show() + else: + mesh = uv2mesh(condition).numpy() + fig = make_subplots(rows=1, cols=condition.shape[0], start_cell="top-left", specs = [ [{ 'type': 'scene' } for _ in range(condition.shape[0])] ]) + F = np.load("custom_utils/smpl_faces.npy") + if is_body: + faces = [F for _ in range(condition.shape[0])] + else: + _, faces = maskFacesBatch(np.ones((condition.shape[0], 14475, 3)), F, mesh_mask) + for i, m in enumerate(mesh): + if not is_body: + m = m[mesh_mask[i], :] + if is_body: + x = m.copy() + m[:,1] = x[:,2] + m[:,2] = x[:,1] + fig.add_trace(go.Mesh3d( + # 8 vertices of a cube + x=m[:,0], + y=m[:,1], + z=m[:,2], + # i, j and k give the vertices of triangles + i = faces[i][:,0], + j = faces[i][:,1], + k = faces[i][:,2] + ), + row=1, col=i+1) + make_square_4(fig) + fig.show() + + +def plot_results(unposed, posed, body, mesh_mask): + fig = make_subplots(rows=1, cols=unposed.shape[0], start_cell="top-left", specs = [ [{ 'type': 'scene' } for _ in range(unposed.shape[0])] ]) + F = np.load("custom_utils/smpl_faces.npy") + _, faces = maskFacesBatch(np.ones((unposed.shape[0], 14475, 3)), F, mesh_mask) + body_mesh = uv2mesh(body).numpy() + + collision = [np.concatenate((body_mesh[i], posed[i][mesh_mask[i], :]), axis=0) for i in range(unposed.shape[0])] + collision_faces = [np.concatenate((F, faces[i] + 14475), axis=0) for i in range(unposed.shape[0])] + collision_color = [np.concatenate((np.ones_like(body_mesh[i], dtype=np.int32)*100, np.ones_like(unposed[i], dtype=np.int32)*230), axis=0) for i in range(unposed.shape[0])] + + for i, c in enumerate(collision): + fig.add_trace(go.Mesh3d( + x=c[:,0], + y=c[:,1], + z=c[:,2], + i=collision_faces[i][:,0], + j=collision_faces[i][:,1], + k=collision_faces[i][:,2], + vertexcolor=collision_color[i] + ), + row=1, col=i+1) + make_square_4(fig) + fig.show() + + + +def get_noise(fix_noise, nr, latent_dim=128): + eps = tf.random.normal(shape=(1 if fix_noise else nr, latent_dim)) + if fix_noise: + eps = tf.tile(eps, [nr,1]) + return eps + def sample(folder, variable_condition, fix_noise, nr=4): # run = wandb.init(project='vae_plots', name=) - random_static, random_body, random_curv, random_unposed = variable_condition == 'static', variable_condition == 'body', variable_condition == 'curv', variable_condition == 'unposed' - frame_main, frame_rand = get_frame(folder, nr) - unposed, uv_mask, mesh_mask = get_unposed(folder, frame, nr, random_unposed) - static = get_statics(folder, nr, random_static) - body, pose, shape = get_bodies(folder, nr, random_body) - curv = get_curvs(folder, body, static, uv_mask, frame, nr, random_curv, curv_model) + frames = get_frame(folder, nr) + noise = get_noise(fix_noise, nr) + noise_curv = get_noise(fix_noise, nr) + body, pose_shape = get_bodies(folder, frames, random_body) + static, uv_mask, mesh_mask = get_statics(folder, frames, random_static) + unposed, mesh_unposed = get_unposed(folder, frames, random_unposed) + curv = get_curvs(folder, body, static, uv_mask, frames, noise_curv, random_curv, curv_model, use_original=True) - pred = model([unposed, static, body, curv, uv_mask], fix_noise=fix_noise) + pred = model.sample(static, body, curv, uv_mask, noise=noise) mesh_pred = uv2mesh(pred) - gender = tf.cast(tf.greater_equal(shape[:, 10] , 0.5), tf.int32) - shape_offsets = tf.stack([tf.gather(tf.tensordot(shape[i,:10][tf.newaxis, :], smpl.shapedirs[gender[i]], axes=[[1], [2]])[0,:,:], smpl.to_keep, axis=0) for i in range(shape.shape[0])]) + gender = tf.cast(tf.greater_equal(pose_shape[:, -1] , 0.5), tf.int32) + shape_offsets = tf.stack([tf.gather(tf.tensordot(pose_shape[i,72:-1][tf.newaxis, :], smpl.shapedirs[gender[i]], axes=[[1], [2]])[0,:,:], smpl.to_keep, axis=0) for i in range(len(frames))]) + mesh_pred_shaped = (mesh_pred + shape_offsets).numpy() + mesh_posed = smpl.run_garment(pose_shape[:, :72], pose_shape[:, 72:], mesh_pred_shaped).numpy() - mesh_pred_shaped = mesh_pred + shape_offsets + plot_inputs(unposed[0], static[0], body[0], curv[0], mesh_mask[0]) - mesh_posed = smpl.run_garment(pose, shape, mesh_pred_shaped).numpy() - - plot_inputs(unposed[0], static[0], body[0], curv[0]) - - variable_cond = None + variable_cond_data = None if random_static: - variable_cond = static + variable_cond_data = static elif random_body: - variable_cond = body + variable_cond_data = body elif random_curv: - variable_cond = curv + variable_cond_data = curv elif random_unposed: - variable_cond = unposed + variable_cond_data = unposed else: assert False - plot_variable_condition(variable_cond, is_curv=random_curv) + plot_variable_condition(variable_cond_data, mesh_mask, is_curv=random_curv, is_body=random_body) - plot_results(mesh_pred, mesh_posed, body, mesh_mask) + plot_results(mesh_pred_shaped, mesh_posed, body, mesh_mask) print() @@ -256,8 +407,8 @@ if __name__ == '__main__': folder = '/home/workspace/local-data/train_100' folder_main = 'checkpoints/cvae_13' - folder_curv = 'checkpoints/cvae_curv_2' - variable_condition = 'curv' + folder_curv = 'checkpoints/cvae_curv_3' + variable_condition = 'static' batch_size = 4 smpl = SMPL()