import torch import os import sys sys.path.append(os.path.dirname(os.path.dirname(os.path.realpath(__file__)))) from vbench2_beta_long import VBenchLong from datetime import datetime import argparse import json def parse_args(): CUR_DIR = os.path.dirname(os.path.abspath(__file__)) parser = argparse.ArgumentParser(description='VBench', formatter_class=argparse.RawTextHelpFormatter) parser.add_argument( "--output_path", type=str, default='./evaluation_results/', help="output path to save the evaluation results", ) parser.add_argument( "--full_json_dir", type=str, default=f'{CUR_DIR}/VBench_full_info.json', help="path to save the json file that contains the prompt and dimension information", ) parser.add_argument( "--videos_path", type=str, required=True, help="folder that contains the sampled videos", ) parser.add_argument( "--dimension", nargs='+', required=True, help="list of evaluation dimensions, usage: --dimension ", ) parser.add_argument( "--load_ckpt_from_local", type=bool, required=False, help="whether load checkpoints from local default paths (assuming you have downloaded the checkpoints locally", ) parser.add_argument( "--read_frame", type=bool, required=False, help="whether directly read frames, or directly read videos", ) parser.add_argument( "--mode", choices=['custom_input', 'vbench_standard', 'vbench_category', 'long_vbench_standard', 'long_custom_input'], default='vbench_standard', help="""This flags determine the mode of evaluations, choose one of the following: 1. "custom_input": receive input prompt from either --prompt/--prompt_file flags or the filename 2. "vbench_standard": evaluate on standard prompt suite of VBench 3. "vbench_category": evaluate on specific category """, ) parser.add_argument( "--custom_input", action="store_true", required=False, help="(deprecated) use --mode=\"custom_input\" instead", ) parser.add_argument( "--prompt", type=str, default="", help="""Specify the input prompt If not specified, filenames will be used as input prompts * Mutually exclusive to --prompt_file. ** This option must be used with --custom_input flag """ ) parser.add_argument( "--prompt_file", type=str, required=False, help="""Specify the path of the file that contains prompt lists If not specified, filenames will be used as input prompts * Mutually exclusive to --prompt. ** This option must be used with --custom_input flag """ ) parser.add_argument( "--category", type=str, required=False, help="""This is for mode=='vbench_category' The category to evaluate on, usage: --category=animal. """, ) ## for dimension specific params ### parser.add_argument( "--imaging_quality_preprocessing_mode", type=str, required=False, default='longer', help="""This is for setting preprocessing in imaging_quality 1. 'shorter': if the shorter side is more than 512, the image is resized so that the shorter side is 512. 2. 'longer': if the longer side is more than 512, the image is resized so that the longer side is 512. 3. 'shorter_centercrop': if the shorter side is more than 512, the image is resized so that the shorter side is 512. Then the center 512 x 512 after resized is used for evaluation. 4. 'None': no preprocessing """, ) parser.add_argument( "--use_semantic_splitting", action="store_true", required=False, help="""Whether to use semantic splitting tools """, ) # for background consistency's feature extractor models parser.add_argument( "--bg_clip2clip_feat_extractor", type=str, default='dreamsim', choices=['clip', 'dreamsim'], help="""This will select the model to caculate background consistency dimension's scores. """, ) # for subject consistency's feature extractor models parser.add_argument( "--sb_clip2clip_feat_extractor", type=str, default='dinov2', choices=['dino', 'dinov2', 'dreamsim'], help="""This will select the model to caculate subject consistency dimension's scores. """, ) parser.add_argument( "--w_inclip", type=float, default=1.0, help="""Weight for in-clip scores, consistency dimensions """, ) parser.add_argument( "--w_clip2clip", type=float, default=0.0, help="""Weight for clip-clip scores, consistency dimensions """, ) parser.add_argument( "--subject_mapping_file_path", type=str, default=f'{CUR_DIR}/configs/subject_mapping_table.yaml', help="""Mapping table of subject consistency. """, ) parser.add_argument( "--background_mapping_file_path", type=str, default=f'{CUR_DIR}/configs/background_mapping_table.yaml', help="""Mapping table of background consistency. """, ) # Weight params for slow-fast evaluation, subject consistency parser.add_argument( "--slow_fast_eval_config", type=str, default=f'{CUR_DIR}/configs/slow_fast_params.yaml', help="""Config files for different clip length. """, ) # for mixture clip length parser.add_argument( "--clip_length_config", type=str, default='clip_length_mix.yaml', help="""Config files for different clip length. """, ) # for dev branch parser.add_argument( "--dev_flag", action="store_true", help="""Denote the current state of pipeline """, ) # control number of video samples for each prompt parser.add_argument( "--num_of_samples_per_prompt", type=int, default=5, help="""Number of samples for each prompt, i.e. prompt-index.mp4 """, ) # for dev branch parser.add_argument( "--static_filter_flag", action="store_true", help="""Denote the current state of pipeline """, ) args = parser.parse_args() return args def main(): args = parse_args() print(f'args: {args}') device = torch.device("cuda") my_VBench = VBenchLong(device, args.full_json_dir, args.output_path) print(f'start evaluation') current_time = datetime.now().strftime('%Y-%m-%d-%H:%M:%S') kwargs = {} prompt = [] assert args.custom_input == False, "(Deprecated) use --mode=custom_input instead" if (args.prompt_file is not None) and (args.prompt != ""): raise Exception("--prompt_file and --prompt cannot be used together") if (args.prompt_file is not None or args.prompt != "") and (not args.mode=='custom_input'): raise Exception("must set --mode=custom_input for using external prompt") if args.prompt_file: with open(args.prompt_file, 'r') as f: prompt = json.load(f) assert type(prompt) == dict, "Invalid prompt file format. The correct format is {\"video_path\": prompt, ... }" elif args.prompt != "": prompt = [args.prompt] if args.category != "": kwargs['category'] = args.category if not args.dev_flag: args.sb_clip2clip_feat_extractor = 'dino' args.bg_clip2clip_feat_extractor = 'clip' args.w_inclip = 1.0 args.w_clip2clip = 0.0 kwargs['sb_clip2clip_feat_extractor'] = args.sb_clip2clip_feat_extractor kwargs['bg_clip2clip_feat_extractor'] = args.bg_clip2clip_feat_extractor kwargs['imaging_quality_preprocessing_mode'] = args.imaging_quality_preprocessing_mode kwargs['clip_length_config'] = args.clip_length_config kwargs['w_inclip'] = args.w_inclip kwargs['w_clip2clip'] = args.w_clip2clip kwargs['use_semantic_splitting'] = args.use_semantic_splitting kwargs['slow_fast_eval_config'] = args.slow_fast_eval_config kwargs['dev_flag'] = args.dev_flag kwargs['sb_mapping_file_path'] = args.subject_mapping_file_path kwargs['bg_mapping_file_path'] = args.background_mapping_file_path kwargs['num_of_samples_per_prompt'] = args.num_of_samples_per_prompt kwargs['static_filter_flag'] = args.static_filter_flag my_VBench.evaluate( videos_path = args.videos_path, name = f'results_{current_time}', prompt_list=prompt, # pass in [] to read prompt from filename dimension_list = args.dimension, local=args.load_ckpt_from_local, read_frame=args.read_frame, mode=args.mode, **kwargs ) print('done') if __name__ == "__main__": main()