src/create_modules.py

import os
import ffmpeg
from datetime import datetime
from pathlib import Path
import numpy as np
import cv2
import spaces
import shutil
import torch
from omegaconf import OmegaConf
from PIL import Image
from scipy.spatial.transform import Rotation as R
from scipy.interpolate import interp1d
from torchvision import transforms

from diffusers import AutoencoderKL, DDIMScheduler
from omegaconf import OmegaConf
from transformers import CLIPVisionModelWithProjection

from src.models.pose_guider import PoseGuider
from src.models.unet_2d_condition import UNet2DConditionModel
from src.models.unet_3d import UNet3DConditionModel
from src.pipelines.pipeline_pose2vid_long import Pose2VideoPipeline

from src.audio_models.model import Audio2MeshModel
from src.utils.mp_utils  import LMKExtractor
from src.utils.draw_util import FaceMeshVisualizer
from src.utils.util import get_fps, read_frames, save_videos_grid

from src.utils.audio_util import prepare_audio_feature
from src.utils.pose_util import project_points_with_trans, matrix_to_euler_and_translation, euler_and_translation_to_matrix, project_points
from src.utils.crop_face_single import crop_face

class Processer():
    def __init__(self):
        self.a2m_model, self.pipe = self.create_models()
    
    # @spaces.GPU
    def create_models(self):

        config = OmegaConf.load('./configs/prompts/animation_audio.yaml')

        if config.weight_dtype == "fp16":
            weight_dtype = torch.float16
        else:
            weight_dtype = torch.float32
            
        audio_infer_config = OmegaConf.load(config.audio_inference_config)
        # prepare model
        a2m_model = Audio2MeshModel(audio_infer_config['a2m_model'])
        a2m_model.load_state_dict(torch.load(audio_infer_config['pretrained_model']['a2m_ckpt'], map_location="cpu"), strict=False)
        a2m_model.to("cuda").eval()

        vae = AutoencoderKL.from_pretrained(
            config.pretrained_vae_path,
        ).to("cuda", dtype=weight_dtype)

        reference_unet = UNet2DConditionModel.from_pretrained(
            config.pretrained_base_model_path,
            subfolder="unet",
        ).to(dtype=weight_dtype, device="cuda")

        inference_config_path = config.inference_config
        infer_config = OmegaConf.load(inference_config_path)
        denoising_unet = UNet3DConditionModel.from_pretrained_2d(
            config.pretrained_base_model_path,
            config.motion_module_path,
            subfolder="unet",
            unet_additional_kwargs=infer_config.unet_additional_kwargs,
        ).to(dtype=weight_dtype, device="cuda")

        pose_guider = PoseGuider(noise_latent_channels=320, use_ca=True).to(device="cuda", dtype=weight_dtype) # not use cross attention

        image_enc = CLIPVisionModelWithProjection.from_pretrained(
            config.image_encoder_path
        ).to(dtype=weight_dtype, device="cuda")

        sched_kwargs = OmegaConf.to_container(infer_config.noise_scheduler_kwargs)
        scheduler = DDIMScheduler(**sched_kwargs)

        # load pretrained weights
        denoising_unet.load_state_dict(
            torch.load(config.denoising_unet_path, map_location="cpu"),
            strict=False,
        )
        reference_unet.load_state_dict(
            torch.load(config.reference_unet_path, map_location="cpu"),
        )
        pose_guider.load_state_dict(
            torch.load(config.pose_guider_path, map_location="cpu"),
        )

        pipe = Pose2VideoPipeline(
            vae=vae,
            image_encoder=image_enc,
            reference_unet=reference_unet,
            denoising_unet=denoising_unet,
            pose_guider=pose_guider,
            scheduler=scheduler,
        )
        pipe = pipe.to("cuda", dtype=weight_dtype)
        
        return a2m_model, pipe 
        
    
    @spaces.GPU
    def audio2video(self, input_audio, ref_img, headpose_video=None, size=512, steps=25, length=150, seed=42):
        fps = 30
        cfg = 3.5
        
        lmk_extractor = LMKExtractor()
        vis = FaceMeshVisualizer()

        config = OmegaConf.load('./configs/prompts/animation_audio.yaml')
        audio_infer_config = OmegaConf.load(config.audio_inference_config)
        generator = torch.manual_seed(seed)

        width, height = size, size

        date_str = datetime.now().strftime("%Y%m%d")
        time_str = datetime.now().strftime("%H%M")
        save_dir_name = f"{time_str}--seed_{seed}-{size}x{size}"

        save_dir = Path(f"output/{date_str}/{save_dir_name}")
        save_dir.mkdir(exist_ok=True, parents=True)

        ref_image_np = cv2.cvtColor(ref_img, cv2.COLOR_RGB2BGR)
        ref_image_np = crop_face(ref_image_np, lmk_extractor)
        if ref_image_np is None:
            return None, Image.fromarray(ref_img)
        
        ref_image_np = cv2.resize(ref_image_np, (size, size))
        ref_image_pil = Image.fromarray(cv2.cvtColor(ref_image_np, cv2.COLOR_BGR2RGB))
        
        face_result = lmk_extractor(ref_image_np)
        if face_result is None: 
            return None, ref_image_pil
        
        lmks = face_result['lmks'].astype(np.float32)
        ref_pose = vis.draw_landmarks((ref_image_np.shape[1], ref_image_np.shape[0]), lmks, normed=True)
        
        sample = prepare_audio_feature(input_audio, wav2vec_model_path=audio_infer_config['a2m_model']['model_path'])
        sample['audio_feature'] = torch.from_numpy(sample['audio_feature']).float().cuda()
        sample['audio_feature'] = sample['audio_feature'].unsqueeze(0)

        # inference
        pred = self.a2m_model.infer(sample['audio_feature'], sample['seq_len'])
        pred = pred.squeeze().detach().cpu().numpy()
        pred = pred.reshape(pred.shape[0], -1, 3)
        pred = pred + face_result['lmks3d']
        
        if headpose_video is not None:
            pose_seq = get_headpose_temp(headpose_video, lmk_extractor)
        else:
            pose_seq = np.load(config['pose_temp'])
        mirrored_pose_seq = np.concatenate((pose_seq, pose_seq[-2:0:-1]), axis=0)
        cycled_pose_seq = np.tile(mirrored_pose_seq, (sample['seq_len'] // len(mirrored_pose_seq) + 1, 1))[:sample['seq_len']]

        # project 3D mesh to 2D landmark
        projected_vertices = project_points(pred, face_result['trans_mat'], cycled_pose_seq, [height, width])

        pose_images = []
        for i, verts in enumerate(projected_vertices):
            lmk_img = vis.draw_landmarks((width, height), verts, normed=False)
            pose_images.append(lmk_img)

        pose_list = []
        pose_tensor_list = []

        pose_transform = transforms.Compose(
            [transforms.Resize((height, width)), transforms.ToTensor()]
        )
        args_L = len(pose_images) if length==0 or length > len(pose_images) else length
        args_L = min(args_L, 300)
        for pose_image_np in pose_images[: args_L]:
            pose_image_pil = Image.fromarray(cv2.cvtColor(pose_image_np, cv2.COLOR_BGR2RGB))
            pose_tensor_list.append(pose_transform(pose_image_pil))
            pose_image_np = cv2.resize(pose_image_np,  (width, height))
            pose_list.append(pose_image_np)
        
        pose_list = np.array(pose_list)
        
        video_length = len(pose_tensor_list)

        video = self.pipe(
            ref_image_pil,
            pose_list,
            ref_pose,
            width,
            height,
            video_length,
            steps,
            cfg,
            generator=generator,
        ).videos

        save_path = f"{save_dir}/{size}x{size}_{time_str}_noaudio.mp4"
        save_videos_grid(
            video,
            save_path,
            n_rows=1,
            fps=fps,
        )
        
        stream = ffmpeg.input(save_path)
        audio = ffmpeg.input(input_audio)
        ffmpeg.output(stream.video, audio.audio, save_path.replace('_noaudio.mp4', '.mp4'), vcodec='copy', acodec='aac', shortest=None).run()
        os.remove(save_path)
        
        return save_path.replace('_noaudio.mp4', '.mp4'), ref_image_pil
    
    @spaces.GPU
    def video2video(self, ref_img, source_video, size=512, steps=25, length=150, seed=42):
        cfg = 3.5
        
        lmk_extractor = LMKExtractor()
        vis = FaceMeshVisualizer()

        generator = torch.manual_seed(seed)
        width, height = size, size

        date_str = datetime.now().strftime("%Y%m%d")
        time_str = datetime.now().strftime("%H%M")
        save_dir_name = f"{time_str}--seed_{seed}-{size}x{size}"

        save_dir = Path(f"output/{date_str}/{save_dir_name}")
        save_dir.mkdir(exist_ok=True, parents=True)

        ref_image_np = cv2.cvtColor(ref_img, cv2.COLOR_RGB2BGR)
        ref_image_np = crop_face(ref_image_np, lmk_extractor)
        if ref_image_np is None:
            return None, Image.fromarray(ref_img)
        
        ref_image_np = cv2.resize(ref_image_np, (size, size))
        ref_image_pil = Image.fromarray(cv2.cvtColor(ref_image_np, cv2.COLOR_BGR2RGB))
        
        face_result = lmk_extractor(ref_image_np)
        if face_result is None: 
            return None, ref_image_pil
        
        lmks = face_result['lmks'].astype(np.float32)
        ref_pose = vis.draw_landmarks((ref_image_np.shape[1], ref_image_np.shape[0]), lmks, normed=True)

        source_images = read_frames(source_video)
        src_fps = get_fps(source_video)
        pose_transform = transforms.Compose(
            [transforms.Resize((height, width)), transforms.ToTensor()]
        )
        
        step = 1
        if src_fps == 60:
            src_fps = 30
            step = 2
        
        pose_trans_list = []
        verts_list = []
        bs_list = []
        src_tensor_list = []
        args_L = len(source_images) if length==0 or length*step > len(source_images) else length*step
        args_L = min(args_L, 300*step)
        for src_image_pil in source_images[: args_L: step]:
            src_tensor_list.append(pose_transform(src_image_pil))
            src_img_np = cv2.cvtColor(np.array(src_image_pil), cv2.COLOR_RGB2BGR)
            frame_height, frame_width, _ = src_img_np.shape
            src_img_result = lmk_extractor(src_img_np)
            if src_img_result is None:
                break
            pose_trans_list.append(src_img_result['trans_mat'])
            verts_list.append(src_img_result['lmks3d'])
            bs_list.append(src_img_result['bs'])

        trans_mat_arr = np.array(pose_trans_list)
        verts_arr = np.array(verts_list)
        bs_arr = np.array(bs_list)
        min_bs_idx = np.argmin(bs_arr.sum(1))
        
        # compute delta pose
        pose_arr = np.zeros([trans_mat_arr.shape[0], 6])

        for i in range(pose_arr.shape[0]):
            euler_angles, translation_vector = matrix_to_euler_and_translation(trans_mat_arr[i]) # real pose of source
            pose_arr[i, :3] =  euler_angles
            pose_arr[i, 3:6] =  translation_vector
        
        init_tran_vec = face_result['trans_mat'][:3, 3] # init translation of tgt
        pose_arr[:, 3:6] = pose_arr[:, 3:6] - pose_arr[0, 3:6] + init_tran_vec # (relative translation of source) + (init translation of tgt)

        pose_arr_smooth = smooth_pose_seq(pose_arr, window_size=3)
        pose_mat_smooth = [euler_and_translation_to_matrix(pose_arr_smooth[i][:3], pose_arr_smooth[i][3:6]) for i in range(pose_arr_smooth.shape[0])]    
        pose_mat_smooth = np.array(pose_mat_smooth)   

        # face retarget
        verts_arr = verts_arr - verts_arr[min_bs_idx] + face_result['lmks3d']
        # project 3D mesh to 2D landmark
        projected_vertices = project_points_with_trans(verts_arr, pose_mat_smooth, [frame_height, frame_width])
        
        pose_list = []
        for i, verts in enumerate(projected_vertices):
            lmk_img = vis.draw_landmarks((frame_width, frame_height), verts, normed=False)
            pose_image_np = cv2.resize(lmk_img,  (width, height))
            pose_list.append(pose_image_np)
        
        pose_list = np.array(pose_list)
        
        video_length = len(pose_list)

        video = self.pipe(
            ref_image_pil,
            pose_list,
            ref_pose,
            width,
            height,
            video_length,
            steps,
            cfg,
            generator=generator,
        ).videos

        save_path = f"{save_dir}/{size}x{size}_{time_str}_noaudio.mp4"
        save_videos_grid(
            video,
            save_path,
            n_rows=1,
            fps=src_fps,
        )
        
        audio_output = f'{save_dir}/audio_from_video.aac'
        # extract audio
        try:
            ffmpeg.input(source_video).output(audio_output, acodec='copy').run()
            # merge audio and video
            stream = ffmpeg.input(save_path)
            audio = ffmpeg.input(audio_output)
            ffmpeg.output(stream.video, audio.audio, save_path.replace('_noaudio.mp4', '.mp4'), vcodec='copy', acodec='aac', shortest=None).run()
        
            os.remove(save_path)
            os.remove(audio_output)
        except:
            shutil.move(
                save_path,
                save_path.replace('_noaudio.mp4', '.mp4')
            )
        
        return save_path.replace('_noaudio.mp4', '.mp4'), ref_image_pil


def matrix_to_euler_and_translation(matrix):
    rotation_matrix = matrix[:3, :3]
    translation_vector = matrix[:3, 3]
    rotation = R.from_matrix(rotation_matrix)
    euler_angles = rotation.as_euler('xyz', degrees=True)
    return euler_angles, translation_vector


def smooth_pose_seq(pose_seq, window_size=5):
    smoothed_pose_seq = np.zeros_like(pose_seq)

    for i in range(len(pose_seq)):
        start = max(0, i - window_size // 2)
        end = min(len(pose_seq), i + window_size // 2 + 1)
        smoothed_pose_seq[i] = np.mean(pose_seq[start:end], axis=0)

    return smoothed_pose_seq

def get_headpose_temp(input_video, lmk_extractor):
    cap = cv2.VideoCapture(input_video)

    total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
    fps = cap.get(cv2.CAP_PROP_FPS)  

    trans_mat_list = []
    while cap.isOpened():
        ret, frame = cap.read()
        if not ret:
            break

        result = lmk_extractor(frame)
        trans_mat_list.append(result['trans_mat'].astype(np.float32))
    cap.release()

    trans_mat_arr = np.array(trans_mat_list)

    # compute delta pose
    trans_mat_inv_frame_0 = np.linalg.inv(trans_mat_arr[0])
    pose_arr = np.zeros([trans_mat_arr.shape[0], 6])

    for i in range(pose_arr.shape[0]):
        pose_mat = trans_mat_inv_frame_0 @ trans_mat_arr[i]
        euler_angles, translation_vector = matrix_to_euler_and_translation(pose_mat)
        pose_arr[i, :3] =  euler_angles
        pose_arr[i, 3:6] =  translation_vector

    # interpolate to 30 fps
    new_fps = 30
    old_time = np.linspace(0, total_frames / fps, total_frames)
    new_time = np.linspace(0, total_frames / fps, int(total_frames * new_fps / fps))

    pose_arr_interp = np.zeros((len(new_time), 6))
    for i in range(6):
        interp_func = interp1d(old_time, pose_arr[:, i])
        pose_arr_interp[:, i] = interp_func(new_time)

    pose_arr_smooth = smooth_pose_seq(pose_arr_interp)
    
    return pose_arr_smooth