add working version

Caltech-101.py

@@ -18,9 +18,11 @@
 import csv
 import json
 import os
+from pathlib import Path
 
 import datasets
 from datasets.tasks import ImageClassification
+import numpy as np
 
 _CITATION = """\
 @article{FeiFei2004LearningGV,
@@ -43,8 +45,8 @@ _HOMEPAGE = "https://data.caltech.edu/records/20086"
 
 _LICENSE = "CC BY 4.0"
 
-_DATA_URL = "brand_new_data/caltech-101.zip"
-# _DATA_URL = "brand_new_data/caltech-101/101_ObjectCategories.tar.gz"
+# _DATA_URL = "brand_new_data/caltech-101.zip"
+_DATA_URL = "brand_new_data/caltech-101/101_ObjectCategories.tar.gz"
 
 _NAMES = [
     "accordion",
@@ -164,24 +166,21 @@ class Caltech101(datasets.GeneratorBasedBuilder):
             description=_DESCRIPTION,
             features=datasets.Features(
                 {
-                    "img": datasets.Image(),
+                    "image": datasets.Image(),
                     "label": datasets.features.ClassLabel(names=_NAMES),
                 }
             ),
-            supervised_keys=("img", "label"),
+            supervised_keys=("image", "label"),
             homepage=_HOMEPAGE,
             license=_LICENSE,
             citation=_CITATION,
             task_templates=ImageClassification(
-                image_column="img", label_column="label"
+                image_column="image", label_column="label"
             ),
         )
 
     def _split_generators(self, dl_manager):
-        # ----- Work in progress here -----
         data_dir = dl_manager.download_and_extract(_DATA_URL)
-        files = dl_manager.iter_files(data_dir)
-        # ---------------------------------
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
@@ -203,5 +202,32 @@ class Caltech101(datasets.GeneratorBasedBuilder):
 
     # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
     def _generate_examples(self, filepath, split):
-        # TODO
-        pass
+        # Same stratagy as the one proposed in TF datasets 
+        is_train_split = (split == "train")
+        data_dir = Path(filepath) / "101_ObjectCategories"
+        # Sets random seed so the random partitioning of files is the same when
+        # called for the train and test splits.
+        numpy_original_state = np.random.get_state()
+        np.random.seed(1234)
+
+        for class_dir in data_dir.iterdir():
+            fnames = [image_path for image_path in class_dir.iterdir() if image_path.name.endswith(".jpg")]
+            assert [image_path for image_path in class_dir.iterdir() if not image_path.name.endswith(".jpg")] == []
+            # _TRAIN_POINTS_PER_CLASS datapoints are sampled for the train split,
+            # the others constitute the test split.
+            if _TRAIN_POINTS_PER_CLASS > len(fnames):
+                raise ValueError("Fewer than {} ({}) points in class {}".format(
+                    _TRAIN_POINTS_PER_CLASS, len(fnames), class_dir.name))
+            train_fnames = np.random.choice(
+                fnames, _TRAIN_POINTS_PER_CLASS, replace=False)
+            test_fnames = set(fnames).difference(train_fnames)
+            fnames_to_emit = train_fnames if is_train_split else test_fnames
+
+            for image_file in fnames_to_emit:
+                record = {
+                    "image": str(image_file),
+                    "label": class_dir.name.lower(),
+                }
+                yield "%s/%s" % (class_dir.name.lower(), image_file), record
+        # Resets the seeds to their previous states.
+        np.random.set_state(numpy_original_state)

brand_new_data/caltech-101/101_ObjectCategories.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af6ece2f339791ca20f855943d8b55dd60892c0a25105fcd631ee3d6430f9926
+size 131740031

brand_new_data/caltech-101/Annotations.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1717f4e10aa837b05956e3f4c94456527b143eec0d95e935028b30aff40663d8
+size 14028800

brand_new_data/caltech-101/show_annotation.m

@@ -0,0 +1,44 @@
+function show_annotation(imgfile, annotation_file);
+%% 
+%% imgfile: string
+%% annotation_file: string
+%%
+%% written by by Fei-Fei Li - November 2004
+%%
+
+IMTYPE = 'jpg'; 
+GUIDELINE_MODE = 1; 
+%% Parameters 
+%label_abbrev = {'LE', 'RE', 'LN', 'NB', 'RN', 'LM', 'RM'}; 
+LARGEFONT = 28; 
+MEDFONT = 18; 
+BIG_WINDOW = get(0,'ScreenSize'); 
+SMALL_WINDOW = [100 100 512 480]; 
+ 
+%% load the annotated data
+load(annotation_file, 'box_coord', 'obj_contour');
+
+%% Read and display image 
+ima = imread(imgfile); 
+ff=figure(1); clf; imagesc(ima); axis image; axis ij; hold on;
+% black and white images
+if length(size(ima))<3
+   colormap(gray);
+end
+set(ff,'Position',SMALL_WINDOW); 
+   
+%% show box
+box_handle = rectangle('position', [box_coord(3), box_coord(1), box_coord(4)-box_coord(3), box_coord(2)-box_coord(1)]);
+set(box_handle, 'edgecolor','y', 'linewidth',5);
+
+%% show contour
+for cc = 1:size(obj_contour,2)
+   if cc < size(obj_contour,2)
+      plot([obj_contour(1,cc), obj_contour(1,cc+1)]+box_coord(3), [obj_contour(2,cc), obj_contour(2,cc+1)]+box_coord(1), 'r','linewidth',4);
+   else
+      plot([obj_contour(1,cc), obj_contour(1,1)]+box_coord(3), [obj_contour(2,cc), obj_contour(2,1)]+box_coord(1), 'r','linewidth',4);
+   end
+end
+
+title(imgfile);
+