Add tile catalog to region catalog conversion script (#935)

* Add tile catalog to region catalog conversion script

* Fix import

case_studies/adaptive_tiling/convert_tilecat_to_regioncat.py

@@ -0,0 +1,51 @@
+import argparse
+import warnings
+from pathlib import Path
+
+import torch
+from torch.utils.data import DataLoader
+from tqdm import tqdm
+
+from bliss.catalog import TileCatalog
+from bliss.generate import itemize_data
+from case_studies.adaptive_tiling.region_catalog import tile_cat_to_region_cat
+
+parser = argparse.ArgumentParser()
+parser.add_argument("input_path", type=str, help="Path to directory containing tile-based data")
+parser.add_argument("output_path", type=str, help="Path to directory to store converted data")
+h = "Overlap to use for conversion. Default is 0.4."
+parser.add_argument("--overlap", required=False, type=float, default=0.4, help=h)
+args = parser.parse_args()
+
+OVERLAP_SLEN = args.overlap
+INPUT_PATH = Path(args.input_path)
+OUTPUT_PATH = Path(args.output_path)
+
+if not OUTPUT_PATH.exists():
+    Path.mkdir(OUTPUT_PATH, parents=True)
+
+warnings.filterwarnings("error")  # We want the warning to throw an error to know to skip an image
+
+# Iterate over each data file in input directory
+files = list(INPUT_PATH.glob("dataset_*.pt"))
+for filename in tqdm(files):
+    dataloader = DataLoader(torch.load(filename), batch_size=1)  # get one at a time
+    region_data = []
+    skip_count = 0
+    for batch in dataloader:
+        # Try conversion. If it fails, skip this image
+        try:
+            region_cat = tile_cat_to_region_cat(TileCatalog(4, batch["tile_catalog"]), OVERLAP_SLEN)
+        except:  # pylint:disable=bare-except  # noqa
+            skip_count += 1
+            continue
+
+        batch["tile_catalog"] = region_cat.to_dict()
+        region_data.append(batch)
+
+    print("Skipped {skip_count} out of {len(dataloader)} total images")  # noqa: WPS421
+
+    # Save converted data to new file
+    output_filename = OUTPUT_PATH / filename.parts[-1]
+    with open(output_filename, "wb") as f:
+        torch.save(itemize_data(region_data), f)