optimizer code format (#106)

* fix metric's dataset metainfo is none bug

* fix image size type is List bug

* fix code format no pass

* optimizer code format

edgelab/datasets/__init__.py

@@ -1,29 +1,11 @@
-from .cocodataset import CustomCocoDataset
-from .data_preprocessors import *
-from .fomodataset import FomoDatasets
-from .meter import MeterData
-from .pipelines import *
-from .sensordataset import SensorDataset
-from .speechcommand import Speechcommand
-from .transforms import *
-from .utils.functions import fomo_collate
-from .vocdataset import CustomVocdataset
-from .yolodataset import CustomYOLOv5CocoDataset
-
-__all__ = [
- 'Speechcommand',
- 'MeterData',
- 'AudioAugs',
- 'CustomCocoDataset',
- 'CustomVocdataset',
- 'FomoDatasets',
- 'SensorDataset',
- 'RandomResizedCrop',
- 'fomo_collate',
- 'ETADataPreprocessor',
- 'CustomYOLOv5CocoDataset',
- 'SensorDataPreprocessor',
- 'PackSensorInputs',
- 'LoadSensorFromFile',
- 'Bbox2FomoMask',
-]
+from .cocodataset import * # noqa
+from .data_preprocessors import * # noqa
+from .fomodataset import * # noqa
+from .meter import * # noqa
+from .pipelines import * # noqa
+from .sensordataset import * # noqa
+from .speechcommand import * # noqa
+from .transforms import * # noqa
+from .utils.functions import * # noqa
+from .vocdataset import * # noqa
+from .yolodataset import * # noqa

edgelab/datasets/data_preprocessors/SensorDataPreprocessor.py

@@ -1,8 +1,5 @@
-import json
-from typing import Optional, Union
+from typing import Optional
 
-import numpy as np
-import torch
 from mmcls.models.utils.batch_augments import RandomBatchAugment
 from mmcls.structures import (
  ClsDataSample,
@@ -12,10 +9,8 @@
  stack_batch_scores,
  tensor_split,
 )
-from mmengine.logging import MessageHub
 from mmengine.model.base_model.data_preprocessor import BaseDataPreprocessor
 
-from edgelab.engine.utils.batch_augs import BatchAugs
 from edgelab.registry import MODELS
 
 

edgelab/datasets/data_preprocessors/audio_augs.py

@@ -100,7 +100,6 @@ def __init__(self, fs, p=0.5, fc_lp=None, fc_hp=None):
 
  def __call__(self, sample):
  if random.random() < self.p:
- a = 0.25
  if random.random() < 0.5:
  fc = 0.5 + random.random() * 0.25
  filt = scipy.signal.firwin(self.num_taps, fc, window='hamming')

edgelab/datasets/fomodataset.py

@@ -6,11 +6,8 @@
 import torch
 import torchvision
 from mmdet.datasets.coco import CocoDataset
-from mmengine.dataset.base_dataset import BaseDataset
 from mmengine.registry import DATASETS
 from sklearn.metrics import confusion_matrix
-from torch.utils.data import Dataset
-from torchvision.transforms import ToTensor
 
 from .pipelines.composition import AlbCompose
 
@@ -74,42 +71,8 @@ def __len__(self):
  return len(self.data)
 
  def __getitem____(self, index):
- image, ann = self.data[index]
-
- self.prepare_data(idx=index)
- image = np.asarray(image)
- return self.pipeline()
-
- bboxes = []
- labels = []
- min_hw_pixels = 2
- for annotation in ann:
- # coco annotation specific https://cocodataset.org/#format-data
- x, y, width, height = annotation['bbox'][:4]
- if width == 0:
- width += min_hw_pixels
- if height == 0:
- height += min_hw_pixels
- annotation['bbox'][:4] = [x, y, width, height]
- bboxes.append(annotation['bbox'])
- labels.append(annotation['category_id'])
-
- bboxes = np.array(bboxes)
- labels = np.array(labels)
-
- trans_param = {'image': image, 'bboxes': bboxes, self.bbox_params['label_fields'][0]: labels}
-
- result = self.transform(**trans_param)
- image = result['image']
- bboxes = result['bboxes']
- labels = result[self.bbox_params['label_fields'][0]]
-
- H, W, C = image.shape
- bbl = []
- for bbox, l in zip(bboxes, labels):
- bbl.append([0, l, (bbox[0] + (bbox[2] / 2)) / W, (bbox[1] + (bbox[3] / 2)) / H, bbox[2] / W, bbox[3] / H])
-
- return {'inputs': ToTensor()(image), 'data_samples': torch.from_numpy(np.asarray(bbl))}
+ result = self.prepare_data(idx=index)
+ return result
 
  def get_ann_info(self, idx):
  ann = self.__getitem__[idx]['target']

edgelab/datasets/meter.py

@@ -23,7 +23,7 @@ def calc_angle(x1, y1, x2, y2):
  z = math.sqrt(x * x + y * y)
  try:
  angle = math.acos((z**2 + 1 - (x - 1) ** 2 - y**2) / (2 * z * 1)) / math.pi * 180
- except:
+ except Exception:
  angle = 0
 
  if y < 0:

edgelab/datasets/pipelines/__init__.py

@@ -1,5 +1,5 @@
-from .albu import *
+from .albu import * # noqa
 from .audio_augs import AudioAugs
 from .transforms import Bbox2FomoMask
 
-__all__ = ['AudioAugs', 'RandomResizedCrop', 'Bbox2FomoMask']
+__all__ = ['AudioAugs', 'Bbox2FomoMask']

edgelab/datasets/pipelines/audio_augs.py

@@ -100,7 +100,6 @@ def __init__(self, fs, p=0.5, fc_lp=None, fc_hp=None):
 
  def __call__(self, sample):
  if random.random() < self.p:
- a = 0.25
  if random.random() < 0.5:
  fc = 0.5 + random.random() * 0.25
  filt = scipy.signal.firwin(self.num_taps, fc, window='hamming')

edgelab/datasets/sensordataset.py

@@ -1,9 +1,7 @@
-import glob
 import json
 import os
-from typing import List, Optional, Sequence, Tuple, Union
+from typing import Optional, Union
 
-import numpy as np
 from mmcls.datasets import CustomDataset
 
 from edgelab.registry import DATASETS

edgelab/datasets/speechcommand.py

@@ -106,7 +106,7 @@ def load_meta_file(self, root, f_meta):
  def _get_labels(self, root):
  f_names = glob.glob(root + f'{sep}**{sep}*.wav')
  self.labels = sorted(list(set([f.split(f'{os.path.sep}')[-2] for f in f_names])))
- self.labels = sorted([l for l in self.labels if l in self.words])
+ self.labels = sorted([label for label in self.labels if label in self.words])
 
  def __getitem__(self, index):
  fname = self.audio_files[index]

edgelab/datasets/transforms/formatting.py

@@ -1,13 +1,10 @@
-from collections import defaultdict
 from collections.abc import Sequence
-from functools import partial
 
 import numpy as np
 import torch
 from mmcls.structures import ClsDataSample
 from mmcv.transforms import BaseTransform
 from mmengine.utils import is_str
-from PIL import Image
 
 from edgelab.registry import TRANSFORMS
 

edgelab/datasets/transforms/loading.py

@@ -2,7 +2,6 @@
 import warnings
 from typing import Optional
 
-import mmcv
 import mmengine.fileio as fileio
 import numpy as np
 from mmcv.transforms.base import BaseTransform

edgelab/datasets/utils/audio_augs.py

@@ -98,7 +98,6 @@ def __init__(self, fs, p=0.5, fc_lp=None, fc_hp=None):
 
  def __call__(self, sample):
  if random.random() < self.p:
- a = 0.25
  if random.random() < 0.5:
  fc = 0.5 + random.random() * 0.25
  filt = scipy.signal.firwin(self.num_taps, fc, window='hamming')

edgelab/datasets/utils/download.py

@@ -28,16 +28,14 @@ def defile(files, store_dir):
  res.append(cmd)
 
 
-def download(links: List or AnyStr, store_path: AnyStr or __path__, unzip_dir=None):
+def download(links: List or AnyStr, store_path: AnyStr, unzip_dir=None):
  if isinstance(links, str):
  links = [links]
  os.chdir(store_path)
  if not os.path.exists('download'):
  os.mkdir('download')
  os.chdir('download')
 
- print(links)
- print(store_path)
  for link in links:
  file_name = link.split('/')[-1]
  unzip = check_compress(file_name)

edgelab/datasets/utils/functions.py

@@ -6,7 +6,7 @@
 @FUNCTIONS.register_module()
 def fomo_collate(batch):
  img, label = [x['inputs'] for x in batch], [y['data_samples'] for y in batch]
- for i, l in enumerate(label):
- if l.shape[0] > 0:
- l[:, 0] = i
+ for i, label in enumerate(label):
+ if label.shape[0] > 0:
+ label[:, 0] = i
  return dict(inputs=torch.stack(img), data_samples=[DetDataSample(labels=torch.cat(label, 0))])