mmod2.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

from os.path import (
    basename
  # listdir
)
import os.path

import random
import scipy.io
import datasets
import numpy as np
from PIL import Image

_HOMEPAGE = ""
_CITATION = ""
_DESCRIPTION = ""
_LICENSE = ""
_IMAGES_DIR = "images"
_ANNOTATIONS_DIR = "annotations"
_BASE_URL = "data.zip"
_LABEL_COLORS = [(59,193,246), (222,168,51), (161,78,69)]
_METADATA_URLS = [
    "train.txt"
  , "validation.txt"
]

class Mmod2(datasets.GeneratorBasedBuilder):
  """Food-101 Images dataset"""



  def _info(self):
    return datasets.DatasetInfo(
        description=_DESCRIPTION,
        features=datasets.Features(
          {
            "image": datasets.Image(),
            "label": datasets.Image(),
          }
        ),
        supervised_keys=("image", "label"),
        homepage=_HOMEPAGE,
        citation=_CITATION,
        license=_LICENSE,
        # task_templates=[datasets.ImageClassification(image_column="image", label_column="label")],
    )

  def _split_generators(self, dl_manager):
    # archive_path = dl_manager.download(_BASE_URL)
    print("0000000")
    split_metadata_paths = dl_manager.download(_METADATA_URLS)
    # print(f"{split_metadata_paths = }")
    self.data_dir = dl_manager.download_and_extract(_BASE_URL)
    with open(split_metadata_paths[0], encoding="utf-8") as f:
      train_files = set(f.read().split("\n"))

    with open(split_metadata_paths[1], encoding="utf-8") as f:
      validation_files = set(f.read().split("\n"))

    return [
      datasets.SplitGenerator(
        name=datasets.Split.TRAIN,
        gen_kwargs={
          "split_key": "train",
          "images": train_files,
          "metadata_path": split_metadata_paths[0],
        },
      ),
      datasets.SplitGenerator(
        name=datasets.Split.VALIDATION,
        gen_kwargs={
          "split_key": "validation",
          "images": validation_files,
          "metadata_path": split_metadata_paths[1],
        },
      ),
    ]

  # def _split_generators(self, dl_manager):
  #   archive = dl_manager.download(_BASE_URL)

  #   return [
  #       datasets.SplitGenerator(
  #           name=datasets.Split.TRAIN, gen_kwargs={"images": dl_manager.iter_archive(archive), "split": "train"}
  #       ),
  #       datasets.SplitGenerator(
  #           name=datasets.Split.VALIDATION, gen_kwargs={"images": dl_manager.iter_archive(archive), "split": "validation"}
  #       ),
  #   ]

  def _generate_examples(self, images, metadata_path=None, split_key="train"):
    """Generate images and labels for splits."""
    print("11111111")

    print(f"""
      {images = }
      {metadata_path = }
      {split_key = }
    """)

    # metadata_path = "validation.txt" if split_key == "validation" else "train.txt"
    # metadata_path = datasets.download(metadata_path)
    # with open(metadata_path, encoding="utf-8") as f:
    #   files_to_keep = set(f.read().split("\n"))
      
    for file_name in images:
      # print(f"{file_name = }")
      
      # get file name without extension
      # file_basename = basename(file_path)[: -len(".jpg")]

      # if file_basename in files_to_keep:

      raw_image_file = f"{self.data_dir}/data/{_IMAGES_DIR}/{file_name}.jpg"
      annotation_file = f"{self.data_dir}/data/{_ANNOTATIONS_DIR}/{file_name}.mat"
          
      if os.path.exists(raw_image_file) and os.path.exists(annotation_file):
        image = Image.open(raw_image_file)
        annotation = scipy.io.loadmat(annotation_file)["annotations"][0][0][0]

        print(f"yielding {file_name}")
        yield file_name, {
          "image": image,     #   {"path": file_path, "bytes": file_obj.read()},
          "label": self.decode_labels(image, annotation),
        }


  def decode_labels(self, image: Image.Image, mask, num_images=1, num_classes=3):
    """Decode batch of segmentation masks.
    
    Args:
      mask: result of inference after taking argmax.
      num_images: number of images to decode from the batch.
      num_classes: number of classes to predict (including background).
    
    Returns:
      A batch with num_images RGB images of the same size as the input. 
    """
    h, w = mask.shape
    img = image.copy()
    # img = Image.new('RGB', (len(mask[i, 0]), len(mask[i])))
    pixels = img.load()
    # for point_index, point in enumerate(mask):
    #   px = int(point[0])
    #   py = int(point[1])

    #   # if point_index < len(mask) - 1:
    #   #   next_px = int(mask[point_index+1][0])
    #   #   next_py = int(mask[point_index+1][1])

    #   #   # draw line
    #   #   x_direction = 1 if px < next_px else -1
    #   #   y_direction = 1 if py < next_py else -1
    #   #   for ix in range(px, next_px, x_direction):
    #   #     for iy in range(py, next_py, y_direction):
    #   if px < img.size[0] and py < img.size[1]:
    #     pixels[px, py] = _LABEL_COLORS[0]
    return img