import os import torch import numpy as np def load_templates(): templates_path = "assets/expression_templates_famos" classes_to_load = ["lips_back", "rolling_lips", "mouth_side", "kissing", "high_smile", "mouth_up", "mouth_middle", "mouth_down", "blow_cheeks", "cheeks_in", "jaw", "lips_up"] templates = {} for subject in os.listdir(templates_path): if os.path.isdir(os.path.join(templates_path, subject)): for template in os.listdir(os.path.join(templates_path, subject)): if template.endswith(".mp4"): continue if template not in classes_to_load: continue exps = [] for npy_file in os.listdir(os.path.join(templates_path, subject, template)): params = np.load(os.path.join(templates_path, subject, template, npy_file), allow_pickle=True) exp = params.item()['expression'].squeeze() exps.append(exp) templates[subject+template] = np.array(exps) print('Number of expression templates loaded: ', len(templates.keys())) return templates def tensor_to_image(image_tensor): """Converts a tensor to a numpy image.""" image = image_tensor.permute(1,2,0).cpu().numpy()*255.0 image = np.clip(image, 0, 255) image = image.astype(np.uint8) return image def image_to_tensor(image): """Converts a numpy image to a tensor.""" image = torch.from_numpy(image).permute(2,0,1).float()/255.0 return image def count_parameters(model): return sum(p.numel() for p in model.parameters() if p.requires_grad) def freeze_module(module, module_name=None): for param in module.parameters(): param.requires_grad_(False) module.eval() def unfreeze_module(module, module_name=None): for param in module.parameters(): param.requires_grad_(True) module.train() import cv2 from torchvision.utils import make_grid def batch_draw_keypoints(images, landmarks, color=(255, 255, 255), radius=1): if isinstance(landmarks, torch.Tensor): landmarks = landmarks.cpu().numpy() landmarks = landmarks.copy()*112 + 112 if isinstance(images, torch.Tensor): images = images.cpu().numpy().transpose(0, 2, 3, 1) images = (images * 255).astype('uint8') images = np.ascontiguousarray(images[..., ::-1]) plotted_images = [] for image, landmark in zip(images, landmarks): for point in landmark: image = cv2.circle(image, (int(point[0]), int(point[1])), radius, color, -1) plotted_images.append(image) return plotted_images def make_grid_from_opencv_images(images, nrow=12): """ Create a grid of images from the list of cv2 images in images""" images = np.array(images) images = images[..., ::-1] images = np.array(images) images = torch.from_numpy(images).permute(0, 3, 1, 2).float()/255. grid = make_grid(images, nrow=nrow) return grid