Spaces:
Running
Running
File size: 4,612 Bytes
2463977 a9c27b3 2463977 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 |
import os
import re
import json
import requests
import gradio as gr
import pandas as pd
from tqdm import tqdm
from bs4 import BeautifulSoup
cache_json = 'cv_backbones.json'
def parse_url(url):
response = requests.get(url)
html = response.text
return BeautifulSoup(html, 'html.parser')
def special_type(m_ver):
m_type = re.search('[a-zA-Z]+', m_ver).group(0)
if m_type == 'wide' or m_type == 'resnext':
return 'resnet'
elif m_type == 'swin':
return 'swin_transformer'
elif m_type == 'inception':
return 'googlenet'
return m_type
def info_on_dataset(m_ver, m_type, in1k_span):
url_span = in1k_span.find_next_sibling('span', {'class': 's2'})
size_span = url_span.find_next_sibling('span', {'class': 'mi'})
m_url = str(url_span.text[1:-1])
input_size = int(size_span.text)
m_dict = {
'ver': m_ver,
'type': m_type,
'input_size': input_size,
'url': m_url
}
return m_dict, size_span
def gen_dataframe(url='https://pytorch.org/vision/main/_modules/'):
torch_page = parse_url(url)
article = torch_page.find('article', {'id': 'pytorch-article'})
ul = article.find('ul').find('ul')
in1k_v1, in1k_v2 = [], []
for li in tqdm(ul.find_all('li'), desc='Crawling cv backbone info...'):
name = str(li.text)
if name.__contains__('torchvision.models.') and len(name.split('.')) == 3:
if name.__contains__('_api') or \
name.__contains__('feature_extraction') or \
name.__contains__('maxvit'):
continue
href = li.find('a').get('href')
model_page = parse_url(url + href)
divs = model_page.select('div.viewcode-block')
for div in divs:
div_id = str(div['id'])
if div_id.__contains__('_Weights'):
m_ver = div_id.split('_Weight')[0].lower()
if m_ver.__contains__('swin_v2_'):
continue
m_type = special_type(m_ver)
in1k_v1_span = div.find(
name='span',
attrs={'class': 'n'},
string='IMAGENET1K_V1'
)
if not in1k_v1_span:
continue
m_dict, size_span = info_on_dataset(
m_ver,
m_type,
in1k_v1_span
)
in1k_v1.append(m_dict)
in1k_v2_span = size_span.find_next_sibling(
name='span',
attrs={'class': 'n'},
string='IMAGENET1K_V2'
)
if in1k_v2_span:
m_dict, _ = info_on_dataset(
m_ver,
m_type,
in1k_v2_span
)
in1k_v2.append(m_dict)
dataset = {
'IMAGENET1K_V1': in1k_v1,
'IMAGENET1K_V2': in1k_v2
}
with open('IMAGENET1K_V1.jsonl', 'w', encoding='utf-8') as jsonl_file:
for item in in1k_v1:
jsonl_file.write(json.dumps(item) + '\n')
with open('IMAGENET1K_V2.jsonl', 'w', encoding='utf-8') as jsonl_file:
for item in in1k_v2:
jsonl_file.write(json.dumps(item) + '\n')
return dataset
def inference(subset):
cache_json = f'{subset}.jsonl'
if os.path.exists(cache_json):
with open(cache_json, 'r', encoding='utf-8') as jsonl_file:
dataset = [json.loads(line) for line in jsonl_file]
else:
dataset = gen_dataframe()[subset]
return pd.DataFrame(dataset), cache_json
def sync(subset):
cache_json = f'{subset}.jsonl'
if os.path.exists(cache_json):
os.remove(cache_json)
return None
with gr.Blocks() as demo:
with gr.Row():
subset_opt = gr.Dropdown(
choices=['IMAGENET1K_V1', 'IMAGENET1K_V2'],
value='IMAGENET1K_V1'
)
sync_btn = gr.Button("Clean cache")
dld_file = gr.components.File(label="Download JSON lines")
with gr.Row():
data_frame = gr.Dataframe(
headers=["ver", "type", "input_size", "url"]
)
subset_opt.change(
inference,
inputs=subset_opt,
outputs=[data_frame, dld_file]
)
sync_btn.click(
sync,
inputs=subset_opt,
outputs=dld_file
)
demo.launch(share=True)
|