Meehai commited on
Commit
7b47e95
·
1 Parent(s): 58c45ab

almost done

Browse files
dronescapes_reader/dronescapes_representations.py CHANGED
@@ -93,37 +93,56 @@ class SemanticRepresentation(NpzRepresentation):
93
  new_images[x_argmax == i] = self.color_map[i]
94
  return new_images
95
 
96
- # class SemanticMapper(SemanticRepresentation):
97
- # """Maps one or more semantic segmentations to a final one + a merge fn. Copy-pasta from VRE"""
98
- # def __init__(*args, original_classes: list[list[str]], mapping: list[dict[str, list[str]]],
99
- # color_map: list[tuple[int, int, int]],
100
- # merge_fn: Callable[[list[np.ndarray]], np.ndarray] | None = None, **kwargs):
101
- # super().__init__(*args, classes=list(mapping[0].keys()), color_map=color_map, **kwargs)
102
- # assert len(self.dependencies) >= 1, "No dependencies provided. Need at least one semantic segmentation to map."
103
- # assert isinstance(mapping, list), type(mapping)
104
- # assert len(mapping) == (B := len(self.dependencies)), (len(mapping), B)
105
- # assert (A := len(original_classes)) == len(self.dependencies), (A, B)
106
- # assert all(m.keys() == mapping[0].keys() for m in mapping), [list(m.keys()) for m in mapping]
107
- # assert len(color_map) == len(mapping[0].keys()), (len(color_map), len(mapping[0].keys()))
108
- # self.original_classes = original_classes
109
- # self.mapping = mapping
110
-
111
- # def _make_one(self, semantic_dep_data: np.ndarray, mapping: dict[str, list[str]],
112
- # original_classes: list[str]) -> np.ndarray:
113
- # assert semantic_dep_data.dtype in (np.uint8, np.uint16), semantic_dep_data.dtype
114
- # mapping_ix = {list(mapping.keys()).index(k): [original_classes.index(_v)
115
- # for _v in v] for k, v in mapping.items()}
116
- # flat_mapping = {}
117
- # for k, v in mapping_ix.items():
118
- # for _v in v:
119
- # flat_mapping[_v] = k
120
- # mapped_data = np.vectorize(flat_mapping.get)(semantic_dep_data).astype(np.uint8)
121
- # return mapped_data
122
-
 
 
 
 
 
 
 
 
 
123
 
124
-
125
- _color_map = [[0, 255, 0], [0, 127, 0], [255, 255, 0], [255, 255, 255],
126
- [255, 0, 0], [0, 0, 255], [0, 255, 255], [127, 127, 63]]
 
 
 
 
 
 
 
 
 
 
127
  coco_classes = ["person", "bicycle", "car", "motorcycle", "airplane", "bus", "train", "truck", "boat", "traffic light",
128
  "fire hydrant", "stop sign", "parking meter", "bench", "bird", "cat", "dog", "horse", "sheep", "cow",
129
  "elephant", "bear", "zebra", "giraffe", "backpack", "umbrella", "handbag", "tie", "suitcase", "frisbee",
@@ -184,24 +203,41 @@ mapillary_color_map = [[165, 42, 42], [0, 192, 0], [196, 196, 196], [190, 153, 1
184
  [0, 60, 100], [0, 0, 142], [0, 0, 90], [0, 0, 230], [0, 80, 100], [128, 64, 64], [0, 0, 110],
185
  [0, 0, 70], [0, 0, 192], [32, 32, 32], [120, 10, 10]]
186
 
187
- dronescapes_task_types = { # some pre-baked representations
188
- "rgb": RGBRepresentation("rgb"),
189
- "hsv": HSVRepresentation("hsv", dependencies=["rgb"]),
190
- "edges_dexined": EdgesRepresentation("edges_dexined"),
191
- "edges_gb": EdgesRepresentation("edges_gb"),
192
- "depth_dpt": DepthRepresentation("depth_dpt", min_depth=0, max_depth=0.999),
193
- "depth_sfm_manual202204": DepthRepresentation("depth_sfm_manual202204", min_depth=0, max_depth=300),
194
- "depth_ufo": DepthRepresentation("depth_ufo", min_depth=0, max_depth=1),
195
- "depth_marigold": DepthRepresentation("depth_marigold", min_depth=0, max_depth=1),
196
- "normals_sfm_manual202204": NormalsRepresentation("normals_sfm_manual202204"),
197
- "opticalflow_rife": OpticalFlowRepresentation("opticalflow_rife"),
198
- "semantic_segprop8": SemanticRepresentation("semantic_segprop8", classes=8, color_map=_color_map),
199
- "semantic_mask2former_swin_mapillary_converted":
200
- SemanticRepresentation("semantic_mask2former_swin_mapillary_converted", classes=8, color_map=_color_map),
201
- "semantic_mask2former_coco_47429163_0":
202
- SemanticRepresentation("semantic_mask2former_coco_47429163_0", classes=coco_classes, color_map=coco_color_map),
203
- "semantic_mask2former_mapillary_49189528_0":
204
- SemanticRepresentation("semantic_mask2former_mapillary_49189528_0", classes=mapillary_classes,
205
- color_map=mapillary_color_map),
206
- "softseg_gb": NpzRepresentation("softseg_gb", 3),
207
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
93
  new_images[x_argmax == i] = self.color_map[i]
94
  return new_images
95
 
96
+ class SemanticMapper(SemanticRepresentation):
97
+ """Maps one or more semantic segmentations to a final one + a merge fn. Copy-pasta from VRE"""
98
+ def __init__(self, *args, original_classes: list[list[str]], mapping: list[dict[str, list[str]]],
99
+ color_map: list[tuple[int, int, int]],
100
+ merge_fn: Callable[[list[np.ndarray]], np.ndarray] | None = None, **kwargs):
101
+ super().__init__(*args, classes=list(mapping[0].keys()), color_map=color_map, **kwargs)
102
+ assert len(self.dependencies) >= 1, "No dependencies provided. Need at least one semantic segmentation to map."
103
+ assert isinstance(mapping, list), type(mapping)
104
+ assert len(mapping) == (B := len(self.dependencies)), (len(mapping), B)
105
+ assert (A := len(original_classes)) == len(self.dependencies), (A, B)
106
+ assert all(m.keys() == mapping[0].keys() for m in mapping), [list(m.keys()) for m in mapping]
107
+ assert len(color_map) == len(mapping[0].keys()), (len(color_map), len(mapping[0].keys()))
108
+ self.original_classes = original_classes
109
+ self.mapping = mapping
110
+ self.merge_fn = merge_fn if merge_fn is not None else SemanticMapper._default_merge_fn
111
+
112
+ @staticmethod
113
+ def _default_merge_fn(dep_data: list[np.ndarray]) -> np.ndarray:
114
+ if len(dep_data) > 1:
115
+ raise ValueError(f"default_merge_fn doesnt' work with >1 dependencies: {len(dep_data)}")
116
+ return dep_data[0]
117
+
118
+ def _make_one(self, path: Path, mapping: dict[str, list[str]],
119
+ original_classes: list[str]) -> np.ndarray:
120
+ semantic_dep_data: np.ndarray = NpzRepresentation.load_from_disk(self, path).numpy()
121
+ semantic_dep_data = semantic_dep_data.argmax(-1) if len(semantic_dep_data.shape) == 3 else semantic_dep_data
122
+ assert len(semantic_dep_data.shape) == 2, f"Only argmaxed data supported, got: {semantic_dep_data.shape}"
123
+ assert semantic_dep_data.dtype in (np.uint8, np.uint16), semantic_dep_data.dtype
124
+ mapping_ix = {list(mapping.keys()).index(k): [original_classes.index(_v)
125
+ for _v in v] for k, v in mapping.items()}
126
+ flat_mapping = {}
127
+ for k, v in mapping_ix.items():
128
+ for _v in v:
129
+ flat_mapping[_v] = k
130
+ mapped_data = np.vectorize(flat_mapping.get)(semantic_dep_data).astype(np.uint8)
131
+ return mapped_data
132
 
133
+ def load_from_disk(self, path: Path | list[Path]):
134
+ # note: assuming SemanticRepresentation for all deps. TODO: generic deps.
135
+ paths = [path] if isinstance(path, Path) else path
136
+ assert len(paths) == len(self.dependencies), (len(path), len(self.dependencies))
137
+ individual_semantics = []
138
+ for path, mapping, original_classes in zip(paths, self.mapping, self.original_classes):
139
+ individual_semantics.append(self._make_one(path, mapping, original_classes))
140
+ res = self.merge_fn(individual_semantics)
141
+ res_torch = F.one_hot(tr.from_numpy(res).long(), num_classes=self.n_classes).float()
142
+ return res_torch
143
+
144
+ color_map_8classes = [[0, 255, 0], [0, 127, 0], [255, 255, 0], [255, 255, 255],
145
+ [255, 0, 0], [0, 0, 255], [0, 255, 255], [127, 127, 63]]
146
  coco_classes = ["person", "bicycle", "car", "motorcycle", "airplane", "bus", "train", "truck", "boat", "traffic light",
147
  "fire hydrant", "stop sign", "parking meter", "bench", "bird", "cat", "dog", "horse", "sheep", "cow",
148
  "elephant", "bear", "zebra", "giraffe", "backpack", "umbrella", "handbag", "tie", "suitcase", "frisbee",
 
203
  [0, 60, 100], [0, 0, 142], [0, 0, 90], [0, 0, 230], [0, 80, 100], [128, 64, 64], [0, 0, 110],
204
  [0, 0, 70], [0, 0, 192], [32, 32, 32], [120, 10, 10]]
205
 
206
+ m2f_mapillary_to_8_classes = {
207
+ "land": ["Terrain", "Sand", "Snow"],
208
+ "forest": ["Vegetation"],
209
+ "residential": ["Building", "Utility Pole", "Pole", "Fence", "Wall", "Manhole", "Street Light", "Curb",
210
+ "Guard Rail", "Caravan", "Junction Box", "Traffic Sign (Front)", "Billboard", "Banner",
211
+ "Mailbox", "Traffic Sign (Back)", "Bench", "Fire Hydrant", "Trash Can", "CCTV Camera",
212
+ "Traffic Light", "Barrier", "Rail Track", "Phone Booth", "Curb Cut", "Traffic Sign Frame",
213
+ "Bike Rack"],
214
+ "road": ["Road", "Lane Marking - General", "Sidewalk", "Bridge", "Other Vehicle", "Motorcyclist", "Pothole",
215
+ "Catch Basin", "Car Mount", "Tunnel", "Parking", "Service Lane", "Lane Marking - Crosswalk",
216
+ "Pedestrian Area", "On Rails", "Bike Lane", "Crosswalk - Plain"],
217
+ "little-objects": ["Car", "Person", "Truck", "Boat", "Wheeled Slow", "Trailer", "Ground Animal", "Bicycle",
218
+ "Motorcycle", "Bird", "Bus", "Ego Vehicle", "Bicyclist", "Other Rider"],
219
+ "water": ["Water"],
220
+ "sky": ["Sky"],
221
+ "hill": ["Mountain"]
222
+ }
223
+
224
+ tasks = [ # some pre-baked representations
225
+ RGBRepresentation("rgb"),
226
+ HSVRepresentation("hsv", dependencies=["rgb"]),
227
+ EdgesRepresentation("edges_dexined"),
228
+ EdgesRepresentation("edges_gb"),
229
+ DepthRepresentation("depth_dpt", min_depth=0, max_depth=0.999),
230
+ DepthRepresentation("depth_sfm_manual202204", min_depth=0, max_depth=300),
231
+ DepthRepresentation("depth_ufo", min_depth=0, max_depth=1),
232
+ DepthRepresentation("depth_marigold", min_depth=0, max_depth=1),
233
+ NormalsRepresentation("normals_sfm_manual202204"),
234
+ OpticalFlowRepresentation("opticalflow_rife"),
235
+ SemanticRepresentation("semantic_segprop8", classes=8, color_map=color_map_8classes),
236
+ SemanticMapper("semantic_mask2former_swin_mapillary_converted", original_classes=[mapillary_classes],
237
+ mapping=[m2f_mapillary_to_8_classes], color_map=color_map_8classes),
238
+ SemanticRepresentation("semantic_mask2former_coco_47429163_0", classes=coco_classes, color_map=coco_color_map),
239
+ SemanticRepresentation("semantic_mask2former_mapillary_49189528_0", classes=mapillary_classes,
240
+ color_map=mapillary_color_map),
241
+ NpzRepresentation("softseg_gb", 3),
242
+ ]
243
+ dronescapes_task_types: dict[str, NpzRepresentation] = {task.name: task for task in tasks}
dronescapes_reader/multitask_dataset.py CHANGED
@@ -41,7 +41,6 @@ class MultiTaskDataset(Dataset):
41
  - 'drop': Drop the data point if any of the representations is missing.
42
  - 'fill_{none,zero,nan}': Fill the missing data with Nones, zeros or NaNs.
43
  - files_suffix: What suffix to look for when creating the dataset. Valid values: 'npy' or 'npz'.
44
- - files_per_repr_overwrites: A dictionay {src: target} that maps one task to another's data (i.e. {'hsv': 'rgb'})
45
  - cache_task_stats: If set to True, the statistics will be cached at '{path}/.task_statistics.npz'. Can be enabled
46
  using the environmental variable STATS_CACHE=1. Defaults to False.
47
  - batch_size_stats: Controls the batch size during statistics computation. Can be enabled by environmental variable
@@ -211,7 +210,8 @@ class MultiTaskDataset(Dataset):
211
  assert not any(len(x) == 0 for x in in_files.values()), f"{ [k for k, v in in_files.items() if len(v) == 0] }"
212
  return in_files
213
 
214
- def _build_dataset(self, task_types: dict[str, NpzRepresentation], task_names: list[str] | None) -> BuildDatasetTuple:
 
215
  logger.debug(f"Building dataset from: '{self.path}'")
216
  all_npz_files = self._get_all_npz_files()
217
  all_files: dict[str, dict[str, str]] = {k: {_v.name: _v for _v in v} for k, v in all_npz_files.items()}
 
41
  - 'drop': Drop the data point if any of the representations is missing.
42
  - 'fill_{none,zero,nan}': Fill the missing data with Nones, zeros or NaNs.
43
  - files_suffix: What suffix to look for when creating the dataset. Valid values: 'npy' or 'npz'.
 
44
  - cache_task_stats: If set to True, the statistics will be cached at '{path}/.task_statistics.npz'. Can be enabled
45
  using the environmental variable STATS_CACHE=1. Defaults to False.
46
  - batch_size_stats: Controls the batch size during statistics computation. Can be enabled by environmental variable
 
210
  assert not any(len(x) == 0 for x in in_files.values()), f"{ [k for k, v in in_files.items() if len(v) == 0] }"
211
  return in_files
212
 
213
+ def _build_dataset(self, task_types: dict[str, NpzRepresentation],
214
+ task_names: list[str] | None) -> BuildDatasetTuple:
215
  logger.debug(f"Building dataset from: '{self.path}'")
216
  all_npz_files = self._get_all_npz_files()
217
  all_files: dict[str, dict[str, str]] = {k: {_v.name: _v for _v in v} for k, v in all_npz_files.items()}
scripts/dronescapes_viewer.py CHANGED
@@ -11,9 +11,8 @@ import random
11
  def main():
12
  assert len(sys.argv) == 2, f"Usage ./dronescapes_viewer.py /path/to/dataset"
13
  reader = MultiTaskDataset(sys.argv[1], task_names=list(dronescapes_task_types.keys()),
14
- task_types=dronescapes_task_types,
15
- handle_missing_data="fill_nan",
16
- normalization="min_max", cache_task_stats=True)
17
  print(reader)
18
 
19
  print("== Shapes ==")
 
11
  def main():
12
  assert len(sys.argv) == 2, f"Usage ./dronescapes_viewer.py /path/to/dataset"
13
  reader = MultiTaskDataset(sys.argv[1], task_names=list(dronescapes_task_types.keys()),
14
+ task_types=dronescapes_task_types, handle_missing_data="fill_nan",
15
+ normalization="min_max", cache_task_stats=True)
 
16
  print(reader)
17
 
18
  print("== Shapes ==")