scripts/2_compute_alisa_scores.py

# %%
import os
import json
import numpy as np
from tqdm import tqdm
from copy import deepcopy
from argparse import ArgumentParser

cmd_args = True
parser = ArgumentParser()
parser.add_argument('--test_dir', default='./test_webpages', help='the directory of test webpages.')
parser.add_argument('--save_dir', default='./save_results', help='the directory for saving result info jsonl file.')
parser.add_argument('--model_name', default="Qwen2.5-VL-32B-Instruct", help='using the vlms for your inference')

if not cmd_args:
    args = parser.parse_args([]) # You can directly set above parameters in the default.
else:
    args = parser.parse_args()

MODEL_NAME = args.model_name

save_path = os.path.join(args.save_dir, f'{MODEL_NAME}.jsonl')
data_info = args.test_dir

with open(data_info) as f:
    data_info_list = f.readlines()
data_info_list = [json.loads(item) for item in data_info_list]
data_info_dict = {
    item['name']: item for item in data_info_list
}

result_sample = {
    'groupLayoutScore': [],
    'overallScore': [],
    'relativeLayoutScore': [],
    'relativeStyleScore': []
}

ranges_count = [0, 50, 100, 150, 200, 400]
# ranges_count = [0, 200, 400]

ranges = {
    key: {
        'data': deepcopy(result_sample),
        'industries': {}
    } for key in ranges_count
}
industries = {}

with open(save_path) as f:
    ori_data = f.readlines()

result = deepcopy(result_sample)

ori_data = [json.loads(item) for item in ori_data]
for item in tqdm(ori_data):
    if item['name'] not in data_info_dict:
        continue
    data_info = data_info_dict[item['name']]
    element_count = data_info.get('element_count', 0)
    for key in result:
        if str(item[key]) == 'nan' or item[key] < 0:
            continue
        result[key].append(item[key])

        append_key = None
        for range_key in ranges:
            if element_count > range_key:
                append_key = range_key
        ranges[append_key]['data'][key].append(item[key])
        industry = data_info['industry']
        if industry not in ranges[append_key]['industries']:
            ranges[append_key]['industries'][industry] = deepcopy(result_sample)
        ranges[append_key]['industries'][industry][key].append(item[key])
        if industry not in industries:
            industries[industry] = deepcopy(result_sample)
        industries[industry][key].append(item[key])

for key in result:
    print(key, np.mean(result[key]))

for key in ranges:
    range_item = ranges[key]
    data = range_item['data']
    for metric_key in data:
        print(f'{key} {metric_key} {np.mean(data[metric_key]):.2f}')
    print('---')

for industry_key in industries:
    idu_item = industries[industry_key]
    for metric_key in idu_item:
        print(f'{industry_key} {metric_key} {np.mean(idu_item[metric_key]):.2f}')
    print('---')