Datasets:
Tasks:
Image Classification
Sub-tasks:
multi-class-image-classification
Languages:
English
Size:
1M<n<10M
ArXiv:
License:
fix error of sometimes having ints in cls_to_idx maps
Browse files
ForNet.py
CHANGED
|
@@ -1134,9 +1134,17 @@ class RecombineDataset(Dataset):
|
|
| 1134 |
for in_cls in bg_rat_indices:
|
| 1135 |
if in_cls not in self.cls_to_idx:
|
| 1136 |
self.cls_to_idx[in_cls] = []
|
| 1137 |
-
for
|
| 1138 |
-
if
|
| 1139 |
-
self.cls_to_idx[in_cls].append(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1140 |
except (TypeError, KeyError, OSError):
|
| 1141 |
logger.warning(
|
| 1142 |
f"Could not load background ratio indices from {bg_rat_indices}. Will do pruning and background selection on the fly. This will take more time in the first few epochs"
|
|
@@ -1476,7 +1484,6 @@ class ForNet(datasets.GeneratorBasedBuilder):
|
|
| 1476 |
"bg_rat_idx_file": datasets.Value("string"),
|
| 1477 |
}
|
| 1478 |
),
|
| 1479 |
-
supervised_keys=("image", "label"),
|
| 1480 |
homepage=_HOMEPAGE,
|
| 1481 |
citation=_CITATION,
|
| 1482 |
)
|
|
@@ -1541,6 +1548,9 @@ class ForNet(datasets.GeneratorBasedBuilder):
|
|
| 1541 |
class_to_zipfile[name.split("/")[-2]] = f
|
| 1542 |
file_ending = "pkl" if name.endswith(".pkl") else "pkl.gz"
|
| 1543 |
name_start = "/".join(name.split("/")[:-2])
|
|
|
|
|
|
|
|
|
|
| 1544 |
if len(name_start) > 0:
|
| 1545 |
name_start += "/"
|
| 1546 |
logger.info(f"Loading extra information: {hf_indices}, {fg_bg_ratios}")
|
|
@@ -1629,7 +1639,7 @@ class ForNet(datasets.GeneratorBasedBuilder):
|
|
| 1629 |
in_cls = data["path"].split("/")[0]
|
| 1630 |
if in_cls not in cls_to_idx:
|
| 1631 |
cls_to_idx[in_cls] = []
|
| 1632 |
-
cls_to_idx[in_cls].append(foraug_idx)
|
| 1633 |
yield foraug_idx, data
|
| 1634 |
foraug_idx += 1
|
| 1635 |
tqdm.write(f"Done generating {split} examples. Saving cls_to_idx file at '{cls_to_idx_loc}'.")
|
|
@@ -1793,4 +1803,3 @@ def _zip_loader(
|
|
| 1793 |
"fg/bg_area": fg_bg_ratios[patch_name],
|
| 1794 |
}
|
| 1795 |
)
|
| 1796 |
-
|
|
|
|
| 1134 |
for in_cls in bg_rat_indices:
|
| 1135 |
if in_cls not in self.cls_to_idx:
|
| 1136 |
self.cls_to_idx[in_cls] = []
|
| 1137 |
+
for data in bg_rat_indices[in_cls]:
|
| 1138 |
+
if isinstance(data, int):
|
| 1139 |
+
self.cls_to_idx[in_cls].append(data)
|
| 1140 |
+
elif isinstance(data, list):
|
| 1141 |
+
idx, rat = data
|
| 1142 |
+
if rat < self.pruning_ratio:
|
| 1143 |
+
self.cls_to_idx[in_cls].append(idx)
|
| 1144 |
+
else:
|
| 1145 |
+
raise TypeError(
|
| 1146 |
+
f"expected entries to be [int, float] (or int), but got {data} ({type(data )}"
|
| 1147 |
+
)
|
| 1148 |
except (TypeError, KeyError, OSError):
|
| 1149 |
logger.warning(
|
| 1150 |
f"Could not load background ratio indices from {bg_rat_indices}. Will do pruning and background selection on the fly. This will take more time in the first few epochs"
|
|
|
|
| 1484 |
"bg_rat_idx_file": datasets.Value("string"),
|
| 1485 |
}
|
| 1486 |
),
|
|
|
|
| 1487 |
homepage=_HOMEPAGE,
|
| 1488 |
citation=_CITATION,
|
| 1489 |
)
|
|
|
|
| 1548 |
class_to_zipfile[name.split("/")[-2]] = f
|
| 1549 |
file_ending = "pkl" if name.endswith(".pkl") else "pkl.gz"
|
| 1550 |
name_start = "/".join(name.split("/")[:-2])
|
| 1551 |
+
else:
|
| 1552 |
+
name_start = ""
|
| 1553 |
+
file_ending = "pkl" if patch_files[0].startswith("train") else "pkl.gz"
|
| 1554 |
if len(name_start) > 0:
|
| 1555 |
name_start += "/"
|
| 1556 |
logger.info(f"Loading extra information: {hf_indices}, {fg_bg_ratios}")
|
|
|
|
| 1639 |
in_cls = data["path"].split("/")[0]
|
| 1640 |
if in_cls not in cls_to_idx:
|
| 1641 |
cls_to_idx[in_cls] = []
|
| 1642 |
+
cls_to_idx[in_cls].append((foraug_idx, data["fg/bg_area"]))
|
| 1643 |
yield foraug_idx, data
|
| 1644 |
foraug_idx += 1
|
| 1645 |
tqdm.write(f"Done generating {split} examples. Saving cls_to_idx file at '{cls_to_idx_loc}'.")
|
|
|
|
| 1803 |
"fg/bg_area": fg_bg_ratios[patch_name],
|
| 1804 |
}
|
| 1805 |
)
|
|
|