titipata

import torch
import gradio as gr
import pytube as pt
from transformers import pipeline
from huggingface_hub import model_info

MODEL_NAME = "biodatlab/whisper-th-medium-combined"  # this always needs to stay in line 8 :D sorry for the hackiness
lang = "th"

device = 0 if torch.cuda.is_available() else "cpu"

titipata

titipata

[{'description': 'บริษัทไปรษณีย์ไทย',
  'vertices': [(528, 202), (741, 202), (741, 222), (528, 222)]},
 {'description': 'จํากัด',
  'vertices': [(754, 204), (809, 204), (809, 222), (754, 222)]},
 {'description': 'พุทธมณฑล',
  'vertices': [(562, 240), (686, 237), (686, 254), (562, 257)]},
 {'description': '73170',
  'vertices': [(707, 234), (777, 232), (777, 253), (707, 255)]},
 {'description': 'สาขาที่',
  'vertices': [(494, 261), (578, 261), (578, 291), (494, 291)]},

titipata

titipata

import os
import json
import torch
import deepcut
import numpy as np

# load model and corpus
device = torch.device("cpu")
with open('./thai-song-model.pt', 'rb') as f:

titipata

# -*- coding: utf-8 -*-
"""
A backend service for Penn event calendar
Run the app:

    $ python api.py

This will serve on localhost:5000.
"""
import os

titipata

Thai songs similarity (experiment)

Create affinity matrix from Thai songs' spectogram

import os
import cv2
from skimage.measure import compare_ssim

dir_name = '/path/to/thaisongs/'

titipata

Convert files in MP4 folder to FLAC

Scipt to convert

import os
from glob import glob
import subprocess

MP4_PATH = '~/Desktop/notebooks/mp4/*.m4a' # path to all mp4 files

titipata

Pretrain bin to txt

Function to turn word vector from http://bio.nlplab.org/#word-vectors into text file format so that easily read.

def w2v_binary_to_text(file_path):
    """
    Function to transform binary file of word vector file
    to text file format
    Ref: http://bio.nlplab.org/#word-vectors

titipata

Script for scraping train arrival time

Get list of date from today until 250 days before.

import requests
import time
import datetime

numdays = 250