fixing script

.gitignore

@@ -1 +1,2 @@
-__pycache__
+__pycache__
+other

image-text-demo.py → loading.py

@@ -1,9 +1,15 @@
 import datasets
 
 _CITATION = """\
+@InProceedings{huggingface:dataset,
+title = {Small image-text set},
+author={James Briggs},
+year={2022}
+}
 """
 
 _DESCRIPTION = """\
+Demo dataset for testing or showing image-text capabilities.
 """
 _HOMEPAGE = "https://huggingface.co/datasets/danbrown/testman-dataset"
 
@@ -29,19 +35,13 @@ class ImageSet(datasets.GeneratorBasedBuilder):
         )
 
     def _split_generators(self, dl_manager):
-        images_archive = dl_manager.download(f"{_REPO}/resolve/main/data/images.tar.gz")
+        images_archive = dl_manager.download(f"{_REPO}/resolve/main/images.tar.gz")
         image_iters = dl_manager.iter_archive(images_archive)
-
-
-        # texts_archive = dl_manager.download(f"{_REPO}/resolve/main/data/texts.tar.gz")
-        # text_iters = dl_manager.iter_archive(texts_archive)
-
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 gen_kwargs={
-                    "images": image_iters,
-                    # "texts": text_iters,
+                    "images": image_iters
                 }
             ),
         ]
@@ -50,12 +50,9 @@ class ImageSet(datasets.GeneratorBasedBuilder):
         """ This function returns the examples in the raw (text) form."""
         
         for idx, (filepath, image) in enumerate(images):
-            
-            # get the corresponding text fime and read it
-            # description = next(texts)[1].read().decode('utf-8')
-
+            description = filepath.split('/')[-1][:-4]
+            description = description.replace('_', ' ')
             yield idx, {
                 "image": {"path": filepath, "bytes": image.read()},
-                "text": "oi"
-                # "text": description,
-            }
+                "text": description,
+            }

scripts/create-jsonl.py

@@ -0,0 +1,23 @@
+# from othersets import load_otherset
+import os
+
+
+# load all text files from ./other/texts, return a list of strings
+def load_texts():
+    texts = []
+    for filename in os.listdir('./other/texts'):
+        with open('./other/texts/' + filename, 'r') as f:
+            texts.append(f.read())
+    return texts
+
+# create a jsonl file with the text content on the 'target' column, and the humber in the 'image_relpath '
+def create_jsonl(texts):
+    with open('./other/texts.jsonl', 'w') as f:
+        for i, text in enumerate(texts):
+            f.write('{"image_relpath": "' + str(i+1) + '.png", "target": "' + text + '"}\n')
+
+
+# start the process
+if __name__ == '__main__':
+    texts = load_texts()
+    create_jsonl(texts)

scripts/load.py

@@ -1,5 +1,4 @@
 from datasets import load_dataset
 
 dataset = load_dataset('danbrown/testman-dataset', split='train')
-
-print(dataset)
+print(dataset[0])