diff --git a/configs/ae_curv_1.py b/configs/ae_curv_1.py new file mode 100644 index 0000000..fa8ff62 --- /dev/null +++ b/configs/ae_curv_1.py @@ -0,0 +1,73 @@ +from utils.convnext_utils import CNBlockConfig + +config = dict() + +config['gpu'] = '2' +config['name'] = 'curv_torch_1' +config['notes'] = ' ' +config['tags'] = ['torch', 'curv'] +config['script'] = 'scripts/ae_curv_train.py' + +config['pretrained_model'] = '' + +config['batch_size'] = 4 +config['train_dir'] = '/data-local/127/train' +config['valid_dir'] = '/data-local/127/valid' +config['test_dir'] = '/data-net/dataset/test_ready' +config['train_frames'] = None +config['valid_frames'] = None + +config['resolution'] = 512 +config['is_skirt'] = False +config['dataset_in_memory'] = False +config['is_cape'] = False +config['remove_global_orientation'] = True +config['device'] = 'cuda' + +config['smpl_res'] = 14475 +if config['is_cape']: + config['smpl_res'] = 3638 + +config['latent_dim'] = 128 + +config['bottleneck_scale'] = 2 + +config['cond_ae'] = dict() +config['cond_ae']['device'] = config['device'] +config['cond_ae']['is_label'] = True +config['cond_ae']['latent_dim'] = config['latent_dim'] +config['cond_ae']['kernel_size'] = 7 +config['cond_ae']['bottleneck_size'] = 40 * config['bottleneck_scale'] + +block_setting = [ + CNBlockConfig(16, 32, 3), + CNBlockConfig(32, 64, 3), + CNBlockConfig(64, 128, 9), + CNBlockConfig(128, config['cond_ae']['bottleneck_size'], 9), + CNBlockConfig(config['cond_ae']['bottleneck_size'], None, 3), +] + +block_setting_reverse = [ + CNBlockConfig(config['cond_ae']['bottleneck_size'], 128, 3), + CNBlockConfig(128, 64, 9), + CNBlockConfig(64, 32, 9), + CNBlockConfig(32, 16, 3), + CNBlockConfig(16, 3, 3), + # CNBlockConfig(3, 3, 3), +] + +config['cond_ae']['encoder_block_settings'] = block_setting +config['cond_ae']['decoder_block_settings'] = block_setting_reverse +config['cond_ae']['activation'] = 'tanh' + + +config['lr'] = 0.001 +config['cond_ae']['upscale_output'] = False + +config['signature'] = { 'id', + 'frame_nr', + 'uv_curvature', + 'mesh_curvature', + 'mesh_mask', + 'uv_mask', + } \ No newline at end of file diff --git a/custom_utils/custom_utils.py b/custom_utils/custom_utils.py index 0f718d5..d9b256d 100644 --- a/custom_utils/custom_utils.py +++ b/custom_utils/custom_utils.py @@ -40,7 +40,7 @@ class UV2Mesh(nn.Module): def uv_to_mesh(self, uv): - if uv.shape[1] == 3: + if uv.shape[1] == 3 or uv.shape[1] == 7: uv = torch.permute(uv, [0,2,3,1]) if uv.shape[1] != 512: uv = Resize(uv[:,:,None], size=(512, 256), interpolation=InterpolationMode.BILINEAR) @@ -65,7 +65,7 @@ class UV2Mesh(nn.Module): # p1 = tf.stack((x1, y1, z1), axis=-1) # p2 = tf.stack((x2, y2, z2), axis=-1) - mesh = torch.mean(uv[:, self.uv_pixels[:,:,0], self.uv_pixels[:,:,1]], dim=2) + mesh = torch.mean(uv[:, self.uv_pixels[:,:,0], self.uv_pixels[:,:,1]].float(), dim=2) return mesh diff --git a/data/dataset.py b/data/dataset.py index ef8a87f..0a2db61 100644 --- a/data/dataset.py +++ b/data/dataset.py @@ -186,7 +186,7 @@ class Dataset(Dataset): uv_curvature = uv_curvature['uv_curvature'] else: uv_curvature = uv_curvature['curvature_uv'] - uv_curvature = torch.nn.functional.one_hot(torch.tensor(uv_curvature, device=self.device, dtype=torch.long), 7).to(self.device) + uv_curvature = torch.nn.functional.one_hot(torch.tensor(uv_curvature, device=self.device, dtype=torch.long), 7).float().to(self.device) item['uv_curvature' + res_string] = uv_curvature if 'uv_unposed' + res_string in self.signature: @@ -206,7 +206,7 @@ class Dataset(Dataset): data = { key: (torch.from_numpy(value) if not (type(value) == str or type(value) == torch.Tensor) else value) for key, value in item.items()} for key, value in data.items(): - if key.startswith('uv_') and value.shape[-1] == 3: + if key.startswith('uv_') and (value.shape[-1] == 3 or value.shape[-1] == 7): data[key] = torch.permute(value, [2,0,1]) if self.cache is not None: diff --git a/models/base.py b/models/base.py index 51051bd..bd4b595 100644 --- a/models/base.py +++ b/models/base.py @@ -33,10 +33,10 @@ class VariationalEncoder(nn.Module): class AEEncoder(nn.Module): - def __init__(self, block_setting: CNBlockConfig, kernel_size: int = 7, activation='tanh', device='cuda'): + def __init__(self, block_setting: CNBlockConfig, kernel_size: int = 7, activation='tanh', is_label=False, nr_of_labels=None, device='cuda'): super(AEEncoder, self).__init__() - self.encoder = ConvNeXt(block_setting=block_setting, kernel_size=kernel_size, device=device) + self.encoder = ConvNeXt(block_setting=block_setting, kernel_size=kernel_size, is_label=is_label, nr_of_labels=nr_of_labels, device=device) if activation == 'tanh': self.act = nn.Tanh() elif activation == 'sigmoid': @@ -81,10 +81,10 @@ class Decoder(nn.Module): x = self.decoder(x) if self.is_label: - x = x * mask[:,:,:,None] + x = x * mask[:,None,:,:] - background = torch.ones((mask.shape[0],512,256,1), device=mask.get_device()) * (1 - mask[:,:,:,None]) - x = torch.concat((background, x), axis=-1) + background = torch.ones((mask.shape[0],1,512,256), device=mask.get_device()) * (1 - mask[:,None,:,:]) + x = torch.concat((background, x), dim=1) x = self.softmax(x) if len(inputs) == 2: diff --git a/models/cond_encoder.py b/models/cond_encoder.py index 34ab6ee..66e16d3 100644 --- a/models/cond_encoder.py +++ b/models/cond_encoder.py @@ -8,8 +8,8 @@ class AutoEncoder(nn.Module): def __init__(self, config): super(AutoEncoder, self).__init__() - self.encoder : AEEncoder = AEEncoder(config['encoder_block_settings'], config['kernel_size'], config['activation'], device=config['device']) - self.decoder : Decoder = Decoder(config['decoder_block_settings'], kernel_size=config['kernel_size'], is_label=config['is_label'], device=config['device']) + self.encoder : AEEncoder = AEEncoder(config['encoder_block_settings'], config['kernel_size'], config['activation'], is_label=config['is_label'], device=config['device'], nr_of_labels=7) + self.decoder : Decoder = Decoder(config['decoder_block_settings'], kernel_size=config['kernel_size'], is_label=config['is_label'], device=config['device'], nr_of_classes=6) def forward(self, input): diff --git a/models/convnext_orig.py b/models/convnext_orig.py index db6b5d1..01b0a84 100644 --- a/models/convnext_orig.py +++ b/models/convnext_orig.py @@ -64,6 +64,8 @@ class ConvNeXtBase(nn.Module): norm_layer: Optional[Callable[..., nn.Module]] = None, device: str = 'cuda', kernel_size: int = 7, + is_label: bool = False, + nr_of_labels: Optional[int] = None, ) -> None: super().__init__() @@ -83,7 +85,7 @@ class ConvNeXtBase(nn.Module): # Stem firstconv_output_channels = block_setting[0].input_channels self.stem = Conv2dNormActivation( - 3, + 3 if not is_label else nr_of_labels, firstconv_output_channels, kernel_size=4, stride=4, @@ -234,7 +236,7 @@ def _convnext( if reversed: model = ConvNeXtReverseBase(block_setting, stochastic_depth_prob=stochastic_depth_prob, is_label=is_label, nr_of_labels=nr_of_labels, kernel_size=kernel_size, device=device, **kwargs) else: - model = ConvNeXtBase(block_setting, stochastic_depth_prob=stochastic_depth_prob, kernel_size=kernel_size, device=device, **kwargs) + model = ConvNeXtBase(block_setting, stochastic_depth_prob=stochastic_depth_prob, kernel_size=kernel_size, is_label=is_label, nr_of_labels=nr_of_labels, device=device, **kwargs) return model diff --git a/scripts/ae_curv_train.py b/scripts/ae_curv_train.py new file mode 100644 index 0000000..858b3c7 --- /dev/null +++ b/scripts/ae_curv_train.py @@ -0,0 +1,221 @@ +import os +import sys +sys.path.append(os.path.dirname(os.path.abspath(__file__)) + '/..') + +import argparse +import importlib + +parser = argparse.ArgumentParser(description='') +parser.add_argument('--name', type=str) +parser.add_argument('--config', type=str) +args = vars(parser.parse_args()) + +config_module = importlib.import_module(args['config']) +config = config_module.config +# from configs.ae_curv_1 import config +os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID" +os.environ["CUDA_VISIBLE_DEVICES"] = config['gpu'] + + +import time +import wandb +import random +import numpy as np +import torch +import cv2 +# from torchmetrics import MultiScaleStructuralSimilarityIndexMeasure as MSSSIM +from custom_utils.custom_utils import UV2Mesh, get_summary_html, maskFacesBatch, get_summary_str +import plotly.graph_objects as go +import plotly.express as ex +import torch.optim as optim +import cmapy as cm + +if config['is_skirt']: + from losses.losses import edge_loss_skirt as edge_loss +else: + from losses.losses import edge_loss + +from data.dataset import Dataset +# from torch.utils.data import DataLoader +from custom_utils.custom_utils import plot_3d +from models.cond_encoder import AutoEncoder +from utils.checkpoints import CheckpointSaver, load_model +from custom_utils.custom_utils import Accumulator +from utils.utils import InfiniteDataLoader as DataLoader +from torchmetrics.classification import MulticlassJaccardIndex +from torchmetrics import Dice +from pytorch_msssim import MS_SSIM + + +def init_wandb(): + run = wandb.init(project='ae', name=config['name'], notes=config['notes'], dir='/data-local/127/temp', tags=config['tags']) + artifact = wandb.Artifact('code', type='code') + artifact.add_file('scripts/ae_curv_train.py') + artifact.add_file('losses/losses.py') + artifact.add_file('models/base.py') + artifact.add_file('models/cond_encoder.py') + artifact.add_file('models/convnext.py') + artifact.add_file('data/dataset.py') + config_file_names = [key for key in sys.modules.keys() if key.startswith('configs.')] + for config_file_name in config_file_names: + file_name = '/'.join(config_file_name.split('.')) + '.py' + if os.path.exists(file_name): + artifact.add_file(file_name) + run.log_artifact(artifact) + +uv2mesh = UV2Mesh(res=256, is_skirt=config['is_skirt'], is_smpl_orig=config['is_cape']) + + +def train_step(batch): + model.train() + pred = model([batch['uv_curvature'], batch['uv_mask']]) + gt_dense = torch.argmax(batch['uv_curvature'], dim=1) + + ce_uv_loss = _ce_uv(pred, gt_dense) + mesh_pred = uv2mesh(pred) * batch['mesh_mask'][:,:,None] + mesh_orig = uv2mesh(gt_dense) * batch['mesh_mask'] + ce_mesh_loss = _ce_mesh(torch.transpose(mesh_pred, 1, 2), mesh_orig.long()) + + loss = ce_uv_loss + ce_mesh_loss + + optimizer.zero_grad() + loss.backward() + optimizer.step() + + model.eval() + with torch.inference_mode(): + ms_ssim = torch.mean(_MSSSIM(torch.argmax(pred, dim=1)[:,None,:,:].float(), gt_dense[:,None,:,:].float())) + mean_iou = _mean_iou(pred, gt_dense) + dice = _dice(pred, gt_dense) + + return { 'train/loss': loss, 'train/ce_uv_loss': ce_uv_loss, 'train/ssim': ms_ssim, 'train/mean_iou': mean_iou, + 'train/dice': dice, 'train/ce_mesh_loss': ce_mesh_loss } + + +def valid_step(batch): + model.eval() + + with torch.inference_mode(): + pred = model([batch['uv_curvature'], batch['uv_mask']]) + gt_dense = torch.argmax(batch['uv_curvature'], dim=1) + + ce_uv_loss = _ce_uv(pred, gt_dense) + mesh_pred = uv2mesh(pred) * batch['mesh_mask'][:,:,None] + mesh_orig = uv2mesh(gt_dense) * batch['mesh_mask'] + ce_mesh_loss = _ce_mesh(torch.transpose(mesh_pred, 1, 2), mesh_orig.long()) + + loss = ce_uv_loss + ce_mesh_loss + + ms_ssim = torch.mean(_MSSSIM(torch.argmax(pred, dim=1)[:,None,:,:].float(), gt_dense[:,None,:,:].float())) + mean_iou = _mean_iou(pred, gt_dense) + dice = _dice(pred, gt_dense) + + return { 'valid/loss': loss, 'valid/ce_uv_loss': ce_uv_loss, 'valid/ssim': ms_ssim, 'valid/mean_iou': mean_iou, + 'valid/dice': dice, 'valid/ce_mesh_loss': ce_mesh_loss } + + +def plot_everything(batch: dict, type: str): + model.eval() + pred = model([batch['uv_curvature'], batch['uv_mask']]) + # gt_dense = torch.argmax(batch['uv_curvature'], dim=1) + + # mesh_pred = uv2mesh(pred) * batch['mesh_mask'][:,:,None] + # mesh_orig = uv2mesh(gt_dense) * batch['mesh_mask'] + + orig_img = torch.argmax(batch['uv_curvature'], dim=1) + pred_img = torch.argmax(pred, dim=1) + img = torch.concat((orig_img, pred_img), axis=2).cpu().numpy() + imgs = [wandb.Image((img[i] + 1) / 2.0) for i in range(4)] + + colors = np.array([cm.color('jet', 1 - i) for i in np.linspace(0, 1, 6+1)]) + color_img = np.zeros((img.shape[0], img.shape[1], img.shape[2], 3), np.uint8) + for i in range(7): + color_img[img == i, :] = colors[i] + color_img = [wandb.Image(color_img[i]) for i in range(4)] + + imgs = { type + '/images': [wandb.Image(img[i,:,:,np.newaxis] / 6.0) for i in range(img.shape[0])], type + '/color_images': color_img } + + return imgs + + +def train(): + valid_iterator = iter(valid_dataloader) + + step = 0 + for batch in train_dataloader: + train_loss = train_step(batch) + train_accumulator.update(train_loss) + + if step % 25 == 0: + data_valid = next(valid_iterator) + valid_loss = valid_step(data_valid) + valid_accumulator.update(valid_loss) + + if step % 1000 == 0 or (step < 1000 and step % 100 == 0): + to_log = dict({ **train_loss, **valid_loss, **{ 'current_step': step } }) + if step % 5000 == 0 or (step < 5000 and step % 1000 == 0): + train_figs = plot_everything(batch, 'train') + valid_figs = plot_everything(data_valid, 'valid') + to_log = { **to_log, **train_figs, **valid_figs } + if step == 0: + to_log = { **to_log, **{ 'model/summary': wandb.Html(get_summary_html(model)) } } + + wandb.log(to_log, step=step) + train_accumulator.reset() + valid_accumulator.reset() + print(step) + + if step % 10000 == 0: + ckpt_saver_enc(model.encoder, step, optimizer) + ckpt_saver_dec(model.decoder, step, optimizer) + + step += 1 + + +if __name__ == '__main__': + init_wandb() + # device = "cuda" if torch.cuda.is_available() else "cpu" + + # parser = argparse.ArgumentParser(description='') + # parser.add_argument('--name', type=str) + # args = vars(parser.parse_args()) + + resolution_modifier = '' + if config['resolution'] != 512: + resolution_modifier = '_' + str(config['resolution']) + skirt_modifier = '_skirt' if config['is_skirt'] else '' + + uv_mask_body = np.load('custom_utils/mask_256x512' + ('_skirt' if config['is_skirt'] else '') + '.npy') + uv_mask_body = cv2.dilate(uv_mask_body.astype(np.float32), cv2.getStructuringElement(cv2.MORPH_RECT, (2 * 2 + 1, 2 * 2 + 1), (2, 2)), iterations=2) + uv_mask_body = torch.from_numpy(uv_mask_body).float().to(config['device']) + + dataset_train = Dataset(config['train_dir'], config['signature'], device=config['device'], frames_file=config['train_frames'], use_cache=config['dataset_in_memory']) + dataset_valid = Dataset(config['valid_dir'], config['signature'], device=config['device'], frames_file=config['valid_frames'], use_cache=config['dataset_in_memory']) + train_dataloader = DataLoader(dataset_train, batch_size=4, drop_last=True, shuffle=True) + valid_dataloader = DataLoader(dataset_valid, batch_size=4, drop_last=True, shuffle=True) + train_accumulator = Accumulator() + valid_accumulator = Accumulator() + + model : AutoEncoder = AutoEncoder(config['cond_ae']) + optimizer = optim.Adam(model.parameters(), lr=config['lr']) + + # _MSSSIM = MSSSIM(data_range=2).to(config['device']) + _MSSSIM = MS_SSIM(data_range=2, size_average=True, channel=1) + _mean_iou = MulticlassJaccardIndex(num_classes=7).to(config['device']) + _dice = Dice(average='macro', num_classes=7).to(config['device']) + _ce_uv = torch.nn.CrossEntropyLoss() + _ce_mesh = torch.nn.CrossEntropyLoss() + + if config['pretrained_model'] != '': + load_model(model.encoder, config['pretrained_model'] + '_enc') + load_model(model.decoder, config['pretrained_model'] + '_dec') + + ckpt_saver_enc = CheckpointSaver('checkpoints/iccv/' + config['name'] + '_enc', top_n=5) + ckpt_saver_dec = CheckpointSaver('checkpoints/iccv/' + config['name'] + '_dec', top_n=5) + + run_name = config['name'] + print(f'started training: {run_name}') + train() + + +