from __future__ import annotations import numpy as np import pandas as pd import requests from huggingface_hub.hf_api import SpaceInfo class PaperList: def __init__(self): self.organization_name = 'ECCV2022' self.table = pd.read_csv('papers.csv') self._preprcess_table() self.table_header = ''' Paper Title Authors pdf Session arXiv GitHub HF Spaces HF Models HF Datasets ''' @staticmethod def load_space_info(author: str) -> list[SpaceInfo]: path = 'https://huggingface.co/api/spaces' r = requests.get(path, params={'author': author}) d = r.json() return [SpaceInfo(**x) for x in d] def add_spaces_to_table(self, organization_name: str, df: pd.DataFrame) -> pd.DataFrame: spaces = self.load_space_info(organization_name) name2space = { s.id.split('/')[1].lower(): f'https://huggingface.co/spaces/{s.id}' for s in spaces } df['hf_space'] = df.loc[:, ['hf_space', 'github']].apply( lambda x: x[0] if isinstance(x[0], str) else name2space.get( x[1].split('/')[-1].lower() if isinstance(x[1], str) else '', np.nan), axis=1) return df def _preprcess_table(self) -> None: self.table = self.add_spaces_to_table(self.organization_name, self.table) self.table['title_lowercase'] = self.table.title.str.lower() rows = [] for row in self.table.itertuples(): paper = f'{row.title}' if isinstance( row.url, str) else row.title pdf = f'pdf' if isinstance( row.pdf, str) else '' arxiv = f'arXiv' if isinstance( row.arxiv, str) else '' github = f'GitHub' if isinstance( row.github, str) else '' hf_space = f'Space' if isinstance( row.hf_space, str) else '' hf_model = f'Model' if isinstance( row.hf_model, str) else '' hf_dataset = f'Dataset' if isinstance( row.hf_dataset, str) else '' row = f''' {paper} {row.authors} {pdf} {row.session} {arxiv} {github} {hf_space} {hf_model} {hf_dataset} ''' rows.append(row) self.table['html_table_content'] = rows def render(self, search_query: str, case_sensitive: bool, filter_names: list[str], paper_sessions: list[str]) -> tuple[int, str]: df = self.add_spaces_to_table(self.organization_name, self.table) if search_query: if case_sensitive: df = df[df.title.str.contains(search_query)] else: df = df[df.title_lowercase.str.contains(search_query.lower())] has_arxiv = 'arXiv' in filter_names has_github = 'GitHub' in filter_names has_hf_space = 'HF Space' in filter_names has_hf_model = 'HF Model' in filter_names has_hf_dataset = 'HF Dataset' in filter_names df = self.filter_table(df, has_arxiv, has_github, has_hf_space, has_hf_model, has_hf_dataset, paper_sessions) return len(df), self.to_html(df, self.table_header) @staticmethod def filter_table(df: pd.DataFrame, has_arxiv: bool, has_github: bool, has_hf_space: bool, has_hf_model: bool, has_hf_dataset: bool, paper_sessions: list[str]) -> pd.DataFrame: if has_arxiv: df = df[~df.arxiv.isna()] if has_github: df = df[~df.github.isna()] if has_hf_space: df = df[~df.hf_space.isna()] if has_hf_model: df = df[~df.hf_model.isna()] if has_hf_dataset: df = df[~df.hf_dataset.isna()] df = df[df.session.isin(set(paper_sessions))] return df @staticmethod def to_html(df: pd.DataFrame, table_header: str) -> str: table_data = ''.join(df.html_table_content) html = f''' {table_header} {table_data}
''' return html