import argparse from omegaconf import OmegaConf import os from musepose_inference import MusePoseInference def parse_args(): parser = argparse.ArgumentParser() parser.add_argument("--config", type=str, default="./configs/test_stage.yaml") parser.add_argument("-W", type=int, default=768, help="Width") parser.add_argument("-H", type=int, default=768, help="Height") parser.add_argument("-L", type=int, default=300, help="video frame length") parser.add_argument("-S", type=int, default=48, help="video slice frame number") parser.add_argument("-O", type=int, default=4, help="video slice overlap frame number") parser.add_argument("--cfg", type=float, default=3.5, help="Classifier free guidance") parser.add_argument("--seed", type=int, default=99) parser.add_argument("--steps", type=int, default=20, help="DDIM sampling steps") parser.add_argument("--fps", type=int) parser.add_argument("--weight_dtype", type=str, default="fp16") parser.add_argument('--model_dir', type=str, default=os.path.join("pretrained_weights"), help='Pretrained models directory for MusePose') parser.add_argument('--output_dir', type=str, default=os.path.join("assets", "videos"), help='Output directory for the result') parser.add_argument("--skip", type=int, default=1, help="frame sample rate = (skip+1)") args = parser.parse_args() return args def main(): args = parse_args() config = OmegaConf.load(args.config) musepose_infer = MusePoseInference( model_dir=args.model_dir, output_dir=args.output_dir ) ref_image_path = list(config["test_cases"].keys())[0] pose_video_path = config["test_cases"][ref_image_path][0] output_file_path, output_demo_file_path = musepose_infer.infer_musepose( ref_image_path=ref_image_path, pose_video_path=pose_video_path, weight_dtype=args.weight_dtype, W=args.W, H=args.H, L=args.L, S=args.S, O=args.O, cfg=args.cfg, seed=args.seed, steps=args.steps, fps=args.fps, skip=args.skip ) print(f"{output_file_path} is saved") if __name__ == "__main__": main()