plot

aaa_helpers.txt

@@ -1,10 +1,3 @@
-nohup python cvae_pyramid.py --name="lhunor_pyr_abl_32_64_e2e_r3" > nohup/review/pyr_abl_32_64_e2e_r3.out 2>&1 &
-nohup python cvae_pyramid.py --name="lhunor_pyr_abl_32_64_e2e_r3_pre" > nohup/review/pyr_abl_32_64_e2e_r3_pre.out 2>&1 &
-nohup python cvae_pyramid.py --name="lhunor_pyr_abl_32_64_e2e_r3_fine" > nohup/review/pyr_abl_32_64_e2e_r3_fine.out 2>&1 &
-nohup python evaluate_ae_pyramid.py --name evaluate_pyr_abl_32_64_e2e_r3 > nohup/evaluate/pyr_abl_32_64_e2e_r3.out 2>&1 &
-nohup python calculate_error.py --name error_pyr_abl_32_64_e2e_r3 > nohup/error/pyr_abl_32_64_e2e_r3.out 2>&1 &
-
-
 one final convolutional layer for smoothing, on the final coordinates
 
 skirts!!!!

calculate_error.py

@@ -1,6 +1,6 @@
 import os
 import sys
-from configs.config_abl_pyr_32 import config
+from configs.config_abl_all_post import config
 
 os.environ["CUDA_VISIBLE_DEVICES"] = config['gpu']
 # import tensorflow as tf
@@ -59,13 +59,13 @@ if __name__ == '__main__':
     folder = config['test_dir']
 
     checkpoint_name = config['name']
-    iteration = '200000'
+    iteration = ''
 
     F = np.load("custom_utils/smpl_faces" + skirt_modifier + ".npy").astype(np.int32)
     garment_type_table = np.load('custom_data/garment_type_table_test.npy', allow_pickle=True).item()
 
 
-    output_folder = '/home/workspace/data/results/' + checkpoint_name + '-' + iteration
+    output_folder = '/home/workspace/data/results/post/' + checkpoint_name + '-' + iteration
 
 
     faces_folder = '/home/workspace/data/results/support/faces' + ('_skirt' if config['is_skirt'] else '')

configs/config_abl_overfit_jumpsuit_2.py

@@ -19,7 +19,7 @@ config['only_curvatures'] =           True
 ablation_count = [config['use_skinning'], config['remove_global_orientation'], config['use_curvature'], config['use_extra_losses'], config['use_offsets']].count(False)
 assert ablation_count == 1 or ablation_count == 0
 
-config['gpu'] = '5'
+config['gpu'] = '3'
 config['name'] = 'abl_overfit_jumpsuit_2'
 config['notes'] = 'remove other conditions from input as well'
 

configs/config_abl_pyr_32.py

@@ -1,9 +1,8 @@
 config = dict()
 
 config['gpu'] = '0'
-config['name'] = 'pyr_abl_32_64_e2e_r3_fine_2'
+config['name'] = 'pyr_abl_32_r1'
 config['notes'] = ''
-config['tags'] = ['review_3']
 
 # DON'T FORGET TO RESET SSIM LOSS USAGE
 # DON'T FORGET TO CHECK MODEL LOADING
@@ -22,12 +21,9 @@ config['use_offsets'] =               False
 
 
 config['pyramid'] = dict()
-config['pyramid']['resolutions'] = [32, 64]
+config['pyramid']['resolutions'] = [32]
 config['pyramid']['pretrained_pyramid'] = ''
-config['pyramid']['final_upscale_enhancer'] = 'checkpoints/64_512_2'
-
-config['pretrained_model'] = 'checkpoints/pyr_abl_32_64_e2e_r3_fine'
-config['pretrained_model_iteration'] = ''
+config['pyramid']['final_upscale_enhancer'] = 'checkpoints/32_512_4'
 
 config['train_dir'] = '/home/workspace/local-data/train'
 config['valid_dir'] = '/home/workspace/local-data/valid'
@@ -44,7 +40,7 @@ config['smpl_res'] = 14475
 if config['is_cape']:
   config['smpl_res'] = 3638
 
-config['lr'] = 0.00001
+config['lr'] = 0.0001
 config['starting_balancing_coeff'] = 0.0001
 config['reg_loss_coeff'] = 0.1 # set to 0 if don't want to use edge+collision
 config['avg_garment_path'] = ''
@@ -68,7 +64,7 @@ config['pyramid']['latent_dim'] = config['latent_dim']
 config['pyramid']['add_noise'] = True
 config['pyramid']['noise_amount'] = 0.025
 
-config['pyramid']['loss_weights'] = [2,1,1,1,1]
+config['pyramid']['loss_weights'] = [1,1,1,1,1]
 
 for res in config['pyramid']['resolutions']:
   if res != 512:
@@ -153,7 +149,7 @@ for res in config['pyramid']['resolutions']:
     level['decoder_depths'] = [9, 9]
     level['decoder_dims'] = [256, 256]
     level['avg_garment_path'] = ''
-    level['upscale_enhancer'] = 'checkpoints/32_512_4'#'checkpoints/up_3_d32_64_2'
+    level['upscale_enhancer'] = ''#'checkpoints/up_3_d32_64_2'
     level['upscale_enhancer_freeze'] = True
     config['pyramid']['levels'].append(level)
   else:

curvature_analysis.py

@@ -0,0 +1,283 @@
+import os
+import sys
+os.environ["CUDA_VISIBLE_DEVICES"] = '4'
+
+
+import tqdm
+import argparse
+import random
+import string
+import time
+import numpy as np
+import trimesh as tm
+import cmapy as cm
+import scipy as sp
+import scipy.io as sio
+import plotly.graph_objects as go
+from pyntcloud import PyntCloud
+from scipy.sparse import csr_matrix, coo_matrix
+from scipy.sparse.csgraph import floyd_warshall, dijkstra
+from plotly.subplots import make_subplots
+from custom_utils.custom_utils import _maskFaces
+from postprocessing import fix_collision
+
+from custom_utils.supersmpl_io import readPC2, readPC2Frame, writePC2, readOBJ
+from custom_utils.custom_utils import _maskFaces
+from auto_encoder_curv import dice_dense
+
+def postprocess(mesh, faces, mask=None, body=None, do_smoothing=True):
+    if body is not None:
+        mesh_fixed = fix_collision(mesh, body[np.newaxis,:,:], mask[np.newaxis,:])[0].numpy()[mask, :]
+    else:
+        mesh_fixed = mesh
+
+    if do_smoothing:
+        m = tm.Trimesh(mesh_fixed, faces, process=False, validate=False)
+        outliers = np.array([len(v) for v in m.vertex_neighbors]) == 0
+        outliers = np.nonzero(outliers)[0]
+        outliers_n_neighbours = get_closest_n_points(mesh_fixed, outliers, 2)
+        outliers_faces = np.concatenate((outliers[:, np.newaxis], outliers_n_neighbours), axis=-1)
+        faces_fixed = np.concatenate((faces, outliers_faces), axis=0)
+        m = tm.Trimesh(mesh_fixed, faces_fixed, process=False, validate=False)
+        tm.smoothing.filter_humphrey(m, iterations=1, alpha=0.6, beta=0.4)
+
+        mesh = m.vertices
+    else:
+        mesh = mesh_fixed
+
+    return mesh
+
+def export_to_ply(mesh, file):
+    ply = tm.exchange.ply.export_ply(mesh, encoding='binary', vertex_normal=True, include_attributes=True)
+    output_file = open(file, "wb+")
+    output_file.write(ply)
+    output_file.close()
+
+
+def get_closest_n_points(verts, idx, n):
+    dist = sp.spatial.distance.cdist(verts[idx, :], verts)
+    dist[dist == 0] = 1e9
+    idx_sort = np.argsort(dist, axis=-1)
+    closest = idx_sort[:, :n]
+    return closest
+
+
+def calculate_dist(folder, neighbourhood):
+    out = dict()
+    frames = sorted(os.listdir(os.path.join(folder, 'mesh_posed')))[int(args.start):int(args.end)]
+
+    for frame in tqdm.tqdm(frames):
+        id = frame.split('_')[0]
+        if id in out.keys():
+            continue
+
+        mask = np.load(os.path.join(folder, 'mesh_mask', f'{id}.npy')).astype(bool)
+        k_neighbors = np.argsort(dist[mask, :][:, mask], axis=-1)[:, :neighbourhood] #cloud.get_neighbors(k=neighbourhood)
+        out[id] = k_neighbors
+
+    return out
+
+
+def histedges_equalN(x, nbin):
+    npt = len(x)
+    return np.interp(np.linspace(0, npt, nbin + 1),
+                     np.arange(npt),
+                     np.sort(x))
+
+
+def get_bins(values, nr_of_bins):
+    flattened = []
+    for c in values:
+        flattened.extend(c.flatten())
+    flattened = np.array(flattened)
+    bins = histedges_equalN(flattened, nr_of_bins)
+    return bins
+
+
+def generate_curvatures(v, faces, mask, neighbourhood=25, nr_of_labels=6):
+    # id = frame.split('_')[0]
+    # frame_nr = int(frame.split('_')[1].split('.')[0])
+
+    # mask = np.load(os.path.join(folder, 'mesh_mask', f'{id}.npy')).astype(bool)
+    # _, faces = _maskFaces(np.zeros((14475,3)), F, mask)
+    # v = np.load(os.path.join(folder, 'mesh_posed', f'{id}_{frame_nr}.npz'))['mesh_posed'][mask,:]
+    
+    m = tm.Trimesh(v, faces, process=False, validate=False)
+
+    outliers = np.array([len(v) for v in m.vertex_neighbors]) == 0
+    outliers = np.nonzero(outliers)[0]
+    outliers_n_neighbours = get_closest_n_points(v, outliers, 2)
+    outliers_faces = np.concatenate((outliers[:, np.newaxis], outliers_n_neighbours), axis=-1)
+    faces = np.concatenate((faces, outliers_faces), axis=0)
+
+    verts = v
+
+    m = tm.Trimesh(verts, faces, process=False, validate=False)
+    export_to_ply(m, temp_file)
+
+    cloud = PyntCloud.from_file(temp_file)
+
+    k_neighbors = np.argsort(dist[mask, :][:, mask], axis=-1)[:, :neighbourhood] #cloud.get_neighbors(k=neighbourhood)
+    #k_neighbors = k_neighbors_cache[id] # np.argsort(dist[mask, :][:, mask], axis=-1)[:, :neighbourhood] #cloud.get_neighbors(k=neighbourhood)
+    ev = cloud.add_scalar_field("eigen_values", k_neighbors=k_neighbors)
+    cloud.add_scalar_field("curvature", ev=ev)
+    curvature = np.array(cloud.points["curvature(" + str(neighbourhood+1) + ")"])
+    
+    bins = np.load(f'custom_data/train_bins.npy') # get_bins(curvature, nr_of_labels)
+    labels = np.linspace(0, bins.shape[0] - 2, bins.shape[0] - 1)
+
+    frame_labels = np.zeros((curvature.shape[0]), dtype=np.float32)
+    for j in range(bins.shape[0] - 2):
+        frame_labels[(curvature >= bins[j]) * (curvature < bins[j+1])] = labels[j]
+    frame_labels[curvature >= bins[-2]] = labels[-1]
+       
+    return frame_labels
+
+import tensorflow as tf
+
+def visualize_curvature_labels(folder, results_folder, frame):
+    id = frame.split('_')[0]
+    frame_nr = frame.split('_')[1].split('.')[0]
+
+    mask = np.load(os.path.join(folder, 'mesh_mask', f'{id}.npy')).astype(bool)
+    _, faces = _maskFaces(np.zeros((14475,3)), F, mask)
+    verts = np.load(os.path.join(folder, 'mesh_posed', f'{id}_{frame_nr}.npz'))['mesh_posed'][mask,:]
+    pred = np.load(os.path.join(results_folder, frame + '.npy'))
+    pred = postprocess(pred, faces)
+
+    curvature_labels_gt = generate_curvatures(verts, faces, mask, 25)
+    curvature_labels_pred = generate_curvatures(pred, faces, mask, 25)
+    # dice = dice_dense(curvature_labels_pred[None,:], curvature_labels_gt[None,:])
+    dice = dice_dense(tf.one_hot(curvature_labels_pred, 7)[None,:], tf.one_hot(curvature_labels_gt, 7)[None,:]).numpy()
+    colors_binned_all = []
+
+    for curvature_labels in [curvature_labels_gt, curvature_labels_pred]:
+        colors = np.array([cm.color('jet', 1 - i) for i in np.linspace(0, 1, int(curvature_labels.max()+1))])
+        colors_binned = np.zeros((curvature_labels.shape[0], 3))
+        for i in range(int(curvature_labels.max())+1):
+            # print(np.count_nonzero(curvature_labels == i))
+            colors_binned[curvature_labels == i, :] = colors[i]
+        colors_binned_all.append(colors_binned)
+
+    fig = make_subplots(rows=1, cols=3, start_cell="top-left", specs = [ [{ 'type': 'scene' } for _ in range(3)] ])
+
+    fig.add_trace(go.Mesh3d(
+                # 8 vertices of a cube
+                x=-pred[:,0],
+                y=pred[:,2],
+                z=pred[:,1],
+                # i, j and k give the vertices of triangles
+                i = faces[:,0],
+                j = faces[:,1],
+                k = faces[:,2],
+                vertexcolor=np.ones((verts.shape[0],3))*np.array([255,200,200])[None,:]
+            ),
+            row=1, col=2)
+
+    fig.add_trace(go.Mesh3d(
+                # 8 vertices of a cube
+                x=verts[:,0],
+                y=verts[:,1],
+                z=verts[:,2],
+                # i, j and k give the vertices of triangles
+                i = faces[:,0],
+                j = faces[:,1],
+                k = faces[:,2],
+                vertexcolor=colors_binned_all[0]
+            ),
+            row=1, col=1)
+    
+    fig.add_trace(go.Mesh3d(
+                # 8 vertices of a cube
+                x=-pred[:,0],
+                y=pred[:,2],
+                z=pred[:,1],
+                # i, j and k give the vertices of triangles
+                i = faces[:,0],
+                j = faces[:,1],
+                k = faces[:,2],
+                vertexcolor=colors_binned_all[1],
+                text=f'dice: {dice}'
+            ),
+            row=1, col=3)
+
+    fig.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),
+                    # scene5_aspectmode='manual', scene5_aspectratio=dict(x=1, y=1, z=1),
+                    )
+    helper = dict(xaxis = dict(range=[-1,1], showbackground=False, showticklabels=False, showaxeslabels=False, showspikes=False, visible=False),
+                yaxis = dict(range=[-1,1], showbackground=False, showticklabels=False, showaxeslabels=False, showspikes=False, visible=False),
+                zaxis = dict(range=[-1,1], showbackground=False, showticklabels=False, showaxeslabels=False, showspikes=False, visible=False),)
+        
+    fig.update_layout(scene = helper,
+                        scene2 = helper,
+                        scene3 = helper,
+                        # scene4 = helper,
+                        # scene5 = helper,
+                        )
+
+
+    # meshes = [go.Mesh3d(
+    #             x=x[:,0],
+    #             y=x[:,1],
+    #             z=x[:,2],
+    #             i = faces[:,0],
+    #             j = faces[:,1],
+    #             k = faces[:,2],
+    #             vertexcolor=colors_binned
+    #             )]
+
+    # fig = go.Figure(data=meshes)
+    # fig.update_layout(scene_aspectmode='manual', scene_aspectratio=dict(x=1, y=1, z=1))
+    # fig.update_layout(
+    #     scene = dict(
+    #         xaxis = dict(nticks=4, range=[-1,1],),
+    #                     yaxis = dict(nticks=4, range=[-1,1],),
+    #                     zaxis = dict(nticks=4, range=[-1,1],),)
+    #                     )
+
+    return fig, dice
+
+import wandb
+
+def init_wandb():
+    run = wandb.init(project='plots', name='review_4', dir='/home/workspace/local-data/temp/')
+    artifact = wandb.Artifact('code', type='code')
+    artifact.add_file('curvature_analysis.py')
+    run.log_artifact(artifact)
+
+if __name__ == '__main__':
+    # init_wandb()
+
+    is_skirt = False
+    skirt_modifier = '_skirt' if is_skirt else ''
+    F = np.load("custom_utils/smpl_faces" + skirt_modifier + ".npy").astype(np.int32)
+    # k_neighbours(k=25)
+
+    folder = '/home/workspace/local-data/train'
+    results_folder = '/home/workspace/data/results/post/abl_overfit_jumpsuit_2-'
+
+    frames_file = 'frames_jumpsuit_30k.txt'
+    frames = []
+    with open(os.path.join(folder, frames_file), 'r') as f:
+        lines = f.readlines()
+        lines = [line.strip() for line in lines]
+    frames = lines
+
+    temp_file = '/home/workspace/local-data/temp/' + ''.join(random.choice(string.ascii_lowercase) for i in range(10)) + '.ply'
+    if not is_skirt:
+        dist = np.load('custom_data/dijkstra.npy')
+    else:
+        dist = np.load('custom_data/dijkstra_skirt.npy')
+
+    i = 0
+    while True:
+        frame = np.random.choice(frames)
+        # frame = '03265_193'
+        frame = '03102_49'
+        fig, dice = visualize_curvature_labels(folder, results_folder, frame)
+        wandb.log({f'{frame}-dice:{dice}': fig}, step=i)
+        i += 1
+        print(frame)

custom_utils/custom_utils.py

@@ -45,7 +45,7 @@ class UV2Mesh(tf.keras.Model):
         if uv.shape[1] == 3:
             uv = tf.transpose(uv, [0,2,3,1])
         if uv.shape[1] != 512:
-            uv = tf.image.resize(uv, (512, 256), tf.image.ResizeMethod.BILINEAR)
+            uv = tf.image.resize(uv, (512, 256), tf.image.ResizeMethod.LANCZOS5)
             
         batch_size = uv.shape[0]
 

cvae.py

@@ -1,6 +1,6 @@
 import os
 import sys
-from configs.config_abl_all_post import config
+from configs.config_abl_overfit_jumpsuit_2 import config
 
 os.environ["CUDA_VISIBLE_DEVICES"] = config['gpu']
 import tensorflow as tf
@@ -41,7 +41,7 @@ from losses import collision_loss as collision_penalty
 from custom_utils.custom_utils import batch_to_dict, load_model, plot_3d, UV2Mesh, get_summary_html, maskFacesBatch, _maskFaces, plot_3d_collision
 from dataset import DatasetFactory, StaticDataset
 from custom_utils.smpl import SMPL
-from models.cvae_model import CVAENew as CVAE
+from models.cvae_model import CVAE
 
 
 def init_wandb():

cvae_pyramid.py

@@ -1,6 +1,6 @@
 import os
 import sys
-from configs.config_abl_pyr_32 import config
+from configs.config_abl_pyr_32_skirt import config
 
 os.environ["CUDA_VISIBLE_DEVICES"] = config['gpu']
 import tensorflow as tf
@@ -210,107 +210,26 @@ def plot_3d(mesh_gt, mesh_pred_unposed_levels, faces, mesh_posed, collision, col
     return fig
 
 
-@tf.function
+# @tf.function
 def train_step(batch):
     levels_input, conds = group_inputs_for_levels(batch, config)
 
     with tf.GradientTape() as tape:
         out_levels = model([levels_input, conds], training=True)
 
-        mesh_orig, uv_orig = batch['mesh_unposed'], batch['uv_unposed']
+        if config['use_skinning']:
+            mesh_orig, uv_orig = batch['mesh_unposed'], batch['uv_unposed']
+        else:
+            mesh_orig, uv_orig = batch['mesh_posed'], batch['uv_posed']
 
         output = dict()
         reconstr_loss_total = 0
         kl_total = 0
 
-        for i, res in enumerate(config['pyramid']['resolutions']):
-            # i = -1
-            # res = config['pyramid']['resolutions'][-1]
-            if res != 512:
-                if i == len(config['pyramid']['resolutions']) - 1:
-                    res_modifier = ''
-                else:
-                    res_modifier = '_' + str(res)
-            else:
-                res_modifier = ''
-
-            pred, mean, logvar = out_levels[i]
-
-            mesh_pred = uv2mesh(pred)
-
-            if config['use_skinning']:
-                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_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)
-            else:
-                mesh_pred_posed = mesh_pred
-
-            # is the output already masked?
-
-            e_loss = tf.reduce_mean(edge_loss(mesh_pred, mesh_orig, batch['mesh_mask']))
-            c_loss, c_ratio = collision_penalty(mesh_pred_posed, batch['mesh_body_posed'], batch['mesh_mask'])
-            c_loss, c_ratio = tf.reduce_mean(c_loss), tf.reduce_mean(c_ratio)
-            l2_loss = tf.reduce_mean(1e3 * tf.reduce_mean(tf.reduce_sum(((batch['uv_unposed' + res_modifier] - pred) ** 2), axis=-1), axis=[1, 2]))
-            # ms_ssim = tf.reduce_mean(tf.image.ssim_multiscale(pred + 5, batch['uv_unposed'] + 5, max_val=10, power_factors=(0.5, 0.2, 0.1, 0.1, 0.1)))
-            # ms_ssim = tf.reduce_mean(tf.image.ssim(pred + 1, batch['uv_unposed'] + 1, max_val=3))
-            l2_mesh_loss = tf.reduce_mean(1e3 * tf.reduce_mean(tf.reduce_sum(batch['mesh_mask'][:,:,tf.newaxis] * (batch['mesh_posed'] - mesh_pred_posed) ** 2, axis=-1), axis=-1))
-            n_loss = normal_loss(mesh_pred_posed, batch['mesh_posed'], batch['mesh_mask'], is_skirt=config['is_skirt'])
-            kl = kl_loss(mean, logvar)
-
-            if config['use_extra_losses']:
-                reconstr_loss_total = reconstr_loss_total + config['pyramid']['loss_weights'][i] * (l2_loss + l2_mesh_loss)
-                if not config['is_cape']:
-                    reconstr_loss_total = reconstr_loss_total + config['pyramid']['loss_weights'][i] * (e_loss + c_loss) * config['reg_loss_coeff'] 
-            else:
-                reconstr_loss_total = reconstr_loss_total + config['pyramid']['loss_weights'][i] * (l2_loss)
-
-            # if config['use_norm_loss'] and step > 2000:
-            #     reconstr_loss_total = reconstr_loss_total + n_loss * config['reg_loss_coeff']
-            
-            kl_total += kl
-
-            mean = tf.reduce_mean(mean)
-            logvar = tf.reduce_mean(logvar)
-
-            metric = tf.reduce_sum(tf.math.sqrt(tf.reduce_sum(batch['mesh_mask'][:,:,tf.newaxis] * (mesh_orig - mesh_pred) ** 2, axis=-1))) / tf.reduce_sum(batch['mesh_mask'])
-
-            output = { **output, **{ 'train/edge_loss' + res_modifier: e_loss, 'train/collision_loss' + res_modifier: c_loss,'train/euclid_error' + res_modifier: metric, 
-                                    'train/mean' + res_modifier: mean, 'train/logvar' + res_modifier: logvar, 'train/kl_loss' + res_modifier: kl,
-                                    'train/var' + res_modifier: np.e ** logvar, 'train/l2_loss' + res_modifier: l2_loss, 'train/normal_loss': n_loss,
-                                    'train/euclid_error': metric, 'train/l2_mesh_loss': l2_mesh_loss } } # 'train/ms_ssim' + res_modifier: ms_ssim,
-
-        loss = reconstr_loss_total + balancing_coeff * kl_total
-
-    gradients = tape.gradient(loss, model.trainable_variables)
-    optimizer.apply_gradients(zip(gradients, model.trainable_variables))
-    return { **{ 'train/loss': loss, 'train/reconstr_loss_total': reconstr_loss_total, 'train/kl_total': kl_total }, **output }
-
-
-@tf.function
-def valid_step(batch):
-    levels_input, conds = group_inputs_for_levels(batch, config)
-    out_levels = model([levels_input, conds], training=False)
-
-    mesh_orig, uv_orig = batch['mesh_unposed'], batch['uv_unposed']
-
-    output = dict()
-    reconstr_loss_total = 0
-    kl_total = 0
-
-    for i, res in enumerate(config['pyramid']['resolutions']):
-        # i = -1
-        # res = config['pyramid']['resolutions'][-1]
-        if res != 512:
-            if i == len(config['pyramid']['resolutions']) - 1:
-                res_modifier = ''
-            else:
-                res_modifier = '_' + str(res)
-        else:
-            res_modifier = ''
-
+        #for i, res in enumerate(config['pyramid']['resolutions']):
+        i = -1
+        res = config['pyramid']['resolutions'][-1]
+        res_modifier = '_' + str(res) if res != 512 else ''
         pred, mean, logvar = out_levels[i]
 
         mesh_pred = uv2mesh(pred)
@@ -324,12 +243,13 @@ def valid_step(batch):
             mesh_pred_posed = smpl.run_garment(batch['pose_shape'][:, :72], batch['pose_shape'][:, 72:], mesh_pred_shaped)
         else:
             mesh_pred_posed = mesh_pred
+
         # is the output already masked?
 
         e_loss = tf.reduce_mean(edge_loss(mesh_pred, mesh_orig, batch['mesh_mask']))
         c_loss, c_ratio = collision_penalty(mesh_pred_posed, batch['mesh_body_posed'], batch['mesh_mask'])
         c_loss, c_ratio = tf.reduce_mean(c_loss), tf.reduce_mean(c_ratio)
-        l2_loss = tf.reduce_mean(1e3 * tf.reduce_mean(tf.reduce_sum(((batch['uv_unposed' + res_modifier] - pred) ** 2), axis=-1), axis=[1, 2]))
+        l2_loss = tf.reduce_mean(1e3 * tf.reduce_mean(tf.reduce_sum(((uv_orig - pred) ** 2), axis=-1), axis=[1, 2]))
         # ms_ssim = tf.reduce_mean(tf.image.ssim_multiscale(pred + 5, batch['uv_unposed'] + 5, max_val=10, power_factors=(0.5, 0.2, 0.1, 0.1, 0.1)))
         # ms_ssim = tf.reduce_mean(tf.image.ssim(pred + 1, batch['uv_unposed'] + 1, max_val=3))
         l2_mesh_loss = tf.reduce_mean(1e3 * tf.reduce_mean(tf.reduce_sum(batch['mesh_mask'][:,:,tf.newaxis] * (batch['mesh_posed'] - mesh_pred_posed) ** 2, axis=-1), axis=-1))
@@ -342,10 +262,10 @@ def valid_step(batch):
                 reconstr_loss_total = reconstr_loss_total + config['pyramid']['loss_weights'][i] * (e_loss + c_loss) * config['reg_loss_coeff'] 
         else:
             reconstr_loss_total = reconstr_loss_total + config['pyramid']['loss_weights'][i] * (l2_loss)
-        
+
         # if config['use_norm_loss'] and step > 2000:
         #     reconstr_loss_total = reconstr_loss_total + n_loss * config['reg_loss_coeff']
-            
+        
         kl_total += kl
 
         mean = tf.reduce_mean(mean)
@@ -353,10 +273,83 @@ def valid_step(batch):
 
         metric = tf.reduce_sum(tf.math.sqrt(tf.reduce_sum(batch['mesh_mask'][:,:,tf.newaxis] * (mesh_orig - mesh_pred) ** 2, axis=-1))) / tf.reduce_sum(batch['mesh_mask'])
 
-        output = { **output, **{ 'valid/edge_loss' + res_modifier: e_loss, 'valid/collision_loss' + res_modifier: c_loss,'valid/euclid_error' + res_modifier: metric, 
-                                    'valid/mean' + res_modifier: mean, 'valid/logvar' + res_modifier: logvar, 'valid/kl_loss' + res_modifier: kl,
-                                    'valid/var' + res_modifier: np.e ** logvar, 'valid/l2_loss' + res_modifier: l2_loss, 'valid/normal_loss': n_loss,
-                                    'valid/euclid_error': metric, 'valid/l2_mesh_loss': l2_mesh_loss } } # 'valid/ms_ssim' + res_modifier: ms_ssim,
+        output = { **output, **{ 'train/edge_loss' + res_modifier: e_loss, 'train/collision_loss' + res_modifier: c_loss,'train/euclid_error' + res_modifier: metric, 
+                                 'train/mean' + res_modifier: mean, 'train/logvar' + res_modifier: logvar, 'train/kl_loss' + res_modifier: kl,
+                                 'train/var' + res_modifier: np.e ** logvar, 'train/l2_loss' + res_modifier: l2_loss, 'train/normal_loss': n_loss,
+                                 'train/euclid_error': metric, 'train/l2_mesh_loss': l2_mesh_loss } } # 'train/ms_ssim' + res_modifier: ms_ssim,
+
+        loss = reconstr_loss_total + balancing_coeff * kl_total
+
+    gradients = tape.gradient(loss, model.trainable_variables)
+    optimizer.apply_gradients(zip(gradients, model.trainable_variables))
+    return { **{ 'train/loss': loss, 'train/reconstr_loss_total': reconstr_loss_total, 'train/kl_total': kl_total }, **output }
+
+
+@tf.function
+def valid_step(batch):
+    levels_input, conds = group_inputs_for_levels(batch, config)
+    out_levels = model([levels_input, conds], training=False)
+
+
+    if config['use_skinning']:
+        mesh_orig, uv_orig = batch['mesh_unposed'], batch['uv_unposed']
+    else:
+        mesh_orig, uv_orig = batch['mesh_posed'], batch['uv_posed']
+
+    output = dict()
+    reconstr_loss_total = 0
+    kl_total = 0
+
+    #for i, res in enumerate(config['pyramid']['resolutions']):
+    i = -1
+    res = config['pyramid']['resolutions'][-1]
+    res_modifier = '_' + str(res) if res != 512 else ''
+    pred, mean, logvar = out_levels[i]
+
+    mesh_pred = uv2mesh(pred)
+
+    if config['use_skinning']:
+        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_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)
+    else:
+        mesh_pred_posed = mesh_pred
+    # is the output already masked?
+
+    e_loss = tf.reduce_mean(edge_loss(mesh_pred, mesh_orig, batch['mesh_mask']))
+    c_loss, c_ratio = collision_penalty(mesh_pred_posed, batch['mesh_body_posed'], batch['mesh_mask'])
+    c_loss, c_ratio = tf.reduce_mean(c_loss), tf.reduce_mean(c_ratio)
+    l2_loss = tf.reduce_mean(1e3 * tf.reduce_mean(tf.reduce_sum(((uv_orig - pred) ** 2), axis=-1), axis=[1, 2]))
+    # ms_ssim = tf.reduce_mean(tf.image.ssim_multiscale(pred + 5, batch['uv_unposed'] + 5, max_val=10, power_factors=(0.5, 0.2, 0.1, 0.1, 0.1)))
+    # ms_ssim = tf.reduce_mean(tf.image.ssim(pred + 1, batch['uv_unposed'] + 1, max_val=3))
+    l2_mesh_loss = tf.reduce_mean(1e3 * tf.reduce_mean(tf.reduce_sum(batch['mesh_mask'][:,:,tf.newaxis] * (batch['mesh_posed'] - mesh_pred_posed) ** 2, axis=-1), axis=-1))
+    n_loss = normal_loss(mesh_pred_posed, batch['mesh_posed'], batch['mesh_mask'], is_skirt=config['is_skirt'])
+    kl = kl_loss(mean, logvar)
+
+    if config['use_extra_losses']:
+        reconstr_loss_total = reconstr_loss_total + config['pyramid']['loss_weights'][i] * (l2_loss + l2_mesh_loss)
+        if not config['is_cape']:
+            reconstr_loss_total = reconstr_loss_total + config['pyramid']['loss_weights'][i] * (e_loss + c_loss) * config['reg_loss_coeff'] 
+    else:
+        reconstr_loss_total = reconstr_loss_total + config['pyramid']['loss_weights'][i] * (l2_loss)
+    
+    # if config['use_norm_loss'] and step > 2000:
+    #     reconstr_loss_total = reconstr_loss_total + n_loss * config['reg_loss_coeff']
+        
+    kl_total += kl
+
+    mean = tf.reduce_mean(mean)
+    logvar = tf.reduce_mean(logvar)
+
+    metric = tf.reduce_sum(tf.math.sqrt(tf.reduce_sum(batch['mesh_mask'][:,:,tf.newaxis] * (mesh_orig - mesh_pred) ** 2, axis=-1))) / tf.reduce_sum(batch['mesh_mask'])
+
+    output = { **output, **{ 'valid/edge_loss' + res_modifier: e_loss, 'valid/collision_loss' + res_modifier: c_loss,'valid/euclid_error' + res_modifier: metric, 
+                                 'valid/mean' + res_modifier: mean, 'valid/logvar' + res_modifier: logvar, 'valid/kl_loss' + res_modifier: kl,
+                                'valid/var' + res_modifier: np.e ** logvar, 'valid/l2_loss' + res_modifier: l2_loss, 'valid/normal_loss': n_loss,
+                                'valid/euclid_error': metric, 'valid/l2_mesh_loss': l2_mesh_loss } } # 'valid/ms_ssim' + res_modifier: ms_ssim,
 
     loss = reconstr_loss_total + balancing_coeff * kl_total
 
@@ -740,7 +733,7 @@ def speed_step(levels_input, conds):
 
 
 if __name__ == '__main__':
-    init_wandb()
+    # init_wandb()
     parser = argparse.ArgumentParser(description='')
     parser.add_argument('--name', type=str, default='lhunor')
     args = vars(parser.parse_args())
@@ -807,19 +800,16 @@ if __name__ == '__main__':
         # assert len(config['pyramid']['resolutions']) == 1
         load_model(model, config['pretrained_model'], iteration=config['pretrained_model_iteration'])
 
-    model.static_enc.trainable = False
-    model.body_enc.trainable = False
-    model.curv_enc.trainable = False
-
-    # model.trainable = False
-    # model.cond_encoder_2.trainable = True
-    # if len(config['pyramid']['resolutions']) == 1:
-    # else:
-    #     load_model(model, config['pyramid']['pretrained_pyramid'], iteration=(config['pyramid']['pretrained_pyramid_iteration'] if 'pretrained_pyramid_iteration' in config.keys() else ''))
-    #     model.static_enc.trainable = False
-    #     model.body_enc.trainable = False
-    #     model.curv_enc.trainable = False
-    #     model.cond_encoder.trainable = False
+    if len(config['pyramid']['resolutions']) == 1:
+        model.static_enc.trainable = False
+        model.body_enc.trainable = False
+        model.curv_enc.trainable = False
+    else:
+        load_model(model, config['pyramid']['pretrained_pyramid'], iteration=(config['pyramid']['pretrained_pyramid_iteration'] if 'pretrained_pyramid_iteration' in config.keys() else ''))
+        model.static_enc.trainable = False
+        model.body_enc.trainable = False
+        model.curv_enc.trainable = False
+        model.cond_encoder.trainable = False
 
     # load_model(model.levels[0].upscale_enhancer, config['pyramid']['levels'][0]['upscale_enhancer'])
     # model.levels[0].upscale_enhancer.trainable = False
@@ -832,8 +822,8 @@ if __name__ == '__main__':
         load_model(model.upscale_enhancer, config['pyramid']['final_upscale_enhancer'])
         model.upscale_enhancer.trainable = False
 
-    # for i in range(len(model.levels)-1):
-    #     model.levels[i].trainable = False
+    for i in range(len(model.levels)-1):
+        model.levels[i].trainable = False
 
     # load_model(model.cond_encoder, config['pyramid']['pretrained_pyramid'])
     # load_model(model.levels[0], config['pyramid']['pretrained_pyramid'])

evaluate_ae.py

@@ -1,6 +1,6 @@
 import os
 import sys
-from configs.config_abl_all_post import config
+from configs.config_abl_overfit_jumpsuit_2 import config
 
 os.environ["CUDA_VISIBLE_DEVICES"] = config['gpu']
 import tensorflow as tf
@@ -42,7 +42,7 @@ from losses import collision_loss as collision_penalty
 from custom_utils.custom_utils import batch_to_dict, load_model, plot_3d, UV2Mesh, get_summary_html, maskFacesBatch, _maskFaces, plot_3d_collision
 from dataset import DatasetFactory, StaticDataset
 from custom_utils.smpl import SMPL
-from models.cvae_model import CVAENew as CVAE
+from models.cvae_model import CVAE
 from custom_utils.supersmpl_io import writeOBJ
 # from unif_metric import chamfer_and_score
 
@@ -104,7 +104,7 @@ if __name__ == '__main__':
 
     iteration = ''
 
-    load_model(model, 'checkpoints/post/' + checkpoint_name, iteration=iteration)
+    load_model(model, 'checkpoints/' + checkpoint_name, iteration=iteration)
     # load_model(model, 'checkpoints/' + 'abl_all_120k')
 
     output_folder = '/home/workspace/data/results/post/' + checkpoint_name + '-' + iteration

evaluate_ae_pyramid.py

@@ -1,6 +1,6 @@
 import os
 import sys
-from configs.config_abl_pyr_32 import config
+from configs.config_abl_2_pyr_32_64_128_256_512 import config
 
 os.environ["CUDA_VISIBLE_DEVICES"] = config['gpu']
 import tensorflow as tf
@@ -97,7 +97,7 @@ if __name__ == '__main__':
     ids = [id.split('_')[0] for id in ids]
     ids = set(ids)
 
-    iteration = '200000'
+    iteration = '50000'
 
     load_model(model, 'checkpoints/' + checkpoint_name, iteration=iteration)
     # load_model(model, 'checkpoints/' + 'abl_all_120k')

models/pyramid_level.py

@@ -92,7 +92,7 @@ class PyramidLevel(tf.keras.Model):
         
         # x = x * uv_mask_high[:,:,:,tf.newaxis]
 
-        return x, z, mean, logvar
+        return x, mean, logvar
 
 
 

models/pyramid_model_optimized.py

@@ -78,11 +78,6 @@ class Pyramid(tf.keras.Model):
             load_model(self.upscale_enhancer, config['pyramid']['final_upscale_enhancer'])
             self.upscale_enhancer.trainable = False
 
-        self.cond_encoder_2 = tf.keras.Sequential([tf.keras.layers.Dense(512, activation='relu'),
-                                                 tf.keras.layers.Dense(256, activation='relu'),
-                                                 tf.keras.layers.Dense(128, activation='relu'),
-                                                 tf.keras.layers.Dense(config['pyramid']['latent_dim'], activation='tanh'),])
-
     def enhance(self, low, mask, training):
         x = self.upscale_enhancer([low, mask], training=training)
         return x
@@ -95,54 +90,44 @@ class Pyramid(tf.keras.Model):
         z_cond = tf.concat([z_static, z_body, z_curv], axis=-1)
         z_cond = self.cond_encoder(z_cond, training=training)
 
-        return z_cond, z_static, z_body, z_curv
-
-    def encode_conds_2(self, z, z_static, z_body, z_curv, training):
-        z_cond = tf.concat([z, z_static, z_body, z_curv], axis=-1)
-        z_cond = self.cond_encoder_2(z_cond, training=training)
-
         return z_cond
 
     def call(self, inputs, training):
         levels_input, conds = inputs
         uv_static, uv_body, uv_curv = conds[0], conds[1], conds[2]
 
-        z_cond, z_static, z_body, z_curv = self.encode_conds(uv_static, uv_body, uv_curv, training)
+        z_cond = self.encode_conds(uv_static, uv_body, uv_curv, training)
 
         if self.add_noise == True and training == True:
             z_cond = z_cond + (tf.random.uniform(z_cond.shape, maxval=self.noise_amount * 2) - self.noise_amount)
 
         out_levels = []
-        pred, z, mean, logvar = self.levels[0](levels_input[0], z_cond, training=training)
+        pred, mean, logvar = self.levels[0](levels_input[0], z_cond, training=training)
         if self.use_offsets:
             pred = pred + levels_input[0][2] # uv_body_posed_32
         if 1 != len(self.levels):
             pred = self.levels[0].enhance(pred, levels_input[0][-2], training=training)  # levels_input[i][-2]=uv_mask_res
+            pred = tf.image.resize(pred, (self.levels[1].size, self.levels[1].size//2), tf.image.ResizeMethod.LANCZOS5)
             out_levels.append([pred, mean, logvar])
-            pred = tf.image.resize(pred, (self.levels[1].size, self.levels[1].size//2), tf.image.ResizeMethod.BILINEAR)
             pred = pred * levels_input[1][-2][:,:,:,tf.newaxis]
         else:
             pred = self.enhance(pred, levels_input[0][-2], training=training)
-            pred = tf.image.resize(pred, (512, 256), tf.image.ResizeMethod.BILINEAR)
+            pred = tf.image.resize(pred, (512, 256), tf.image.ResizeMethod.LANCZOS5)
             pred = pred * levels_input[-1][-1][:,:,:,tf.newaxis] # full res uv mask
             out_levels.append([pred, mean, logvar])
 
         for i in range(1, len(self.levels)):
-            input = levels_input[i]
-            input[0] = pred # use previous level output as input for next level
-            z_cond_2 = self.encode_conds_2(z, z_static, z_body, z_curv, training=training)
-
-            pred_level, z, mean, logvar = self.levels[i](input, z_cond_2, training=training)
+            pred_level, mean, logvar = self.levels[i](levels_input[i], z_cond, training=training)
             pred = pred + pred_level
             if i != len(self.levels) - 1:
                 pred = self.levels[i].enhance(pred, levels_input[i][-2], training=training)
                 out_levels.append([pred, mean, logvar])
-                pred = tf.image.resize(pred, (self.levels[i+1].size, self.levels[i+1].size//2), tf.image.ResizeMethod.BILINEAR)
+                pred = tf.image.resize(pred, (self.levels[i+1].size, self.levels[i+1].size//2), tf.image.ResizeMethod.LANCZOS5)
                 pred = pred * levels_input[i+1][-2][:,:,:,tf.newaxis]
             else:
                 if self.upscale_enhancer is not None:
                     pred = self.enhance(pred, levels_input[-1][-2], training=training)
-                    pred = tf.image.resize(pred, (512, 256), tf.image.ResizeMethod.BILINEAR)
+                    pred = tf.image.resize(pred, (512, 256), tf.image.ResizeMethod.LANCZOS5)
                 pred = pred * levels_input[-1][-1][:,:,:,tf.newaxis] # full res uv mask
                 out_levels.append([pred, mean, logvar])