shortery commited on
Commit
391b5e5
1 Parent(s): 71cfd9a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +47 -3
README.md CHANGED
@@ -21,13 +21,13 @@ dataset_info:
21
  dtype: string
22
  splits:
23
  - name: validation
24
- num_bytes: 64018244.0
25
  num_examples: 101
26
  - name: test
27
- num_bytes: 125460818.0
28
  num_examples: 199
29
  download_size: 189448472
30
- dataset_size: 189479062.0
31
  configs:
32
  - config_name: default
33
  data_files:
@@ -35,4 +35,48 @@ configs:
35
  path: data/validation-*
36
  - split: test
37
  path: data/test-*
 
 
 
 
 
 
38
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21
  dtype: string
22
  splits:
23
  - name: validation
24
+ num_bytes: 64018244
25
  num_examples: 101
26
  - name: test
27
+ num_bytes: 125460818
28
  num_examples: 199
29
  download_size: 189448472
30
+ dataset_size: 189479062
31
  configs:
32
  - config_name: default
33
  data_files:
 
35
  path: data/validation-*
36
  - split: test
37
  path: data/test-*
38
+ license: mit
39
+ task_categories:
40
+ - image-to-text
41
+ - object-detection
42
+ size_categories:
43
+ - n<1K
44
  ---
45
+
46
+ # DM codes dataset
47
+
48
+ The dataset contains photos of Data Matrix (DM) codes and their annotations. The photos were taken on an iPhone and annotated manually by a human.
49
+ The annotations contain **text**, which is encoded in the DM code and the pixel coordinates of the DM code vertices.
50
+ The vertices are: **tl** = top left, **tr** = top right, **br** = bottom right, **bl** = bottom left.
51
+ Attribute **is_clean** specifies whether the DM code on the image is expected to be easily readable. For every DM code, there is exactly one image
52
+ with `is_clean=true` and several images with `is_clean=false`.
53
+
54
+ If you want to crop the DM codes from the images, use the following code:
55
+
56
+ ```python
57
+ import numpy as np
58
+ import datasets
59
+ from PIL import Image
60
+ from skimage import transform
61
+
62
+ def crop_dm_code(example: dict, square_side: int = 200, square_padding: int = 25) -> dict:
63
+ vertices = np.asarray((example["tl"], example["tr"], example["br"], example["bl"]))
64
+ unit_square = np.asarray([
65
+ [square_padding, square_padding],
66
+ [square_side + square_padding, square_padding],
67
+ [square_side + square_padding, square_side + square_padding],
68
+ [square_padding, square_side + square_padding]
69
+ ])
70
+ transf = transform.ProjectiveTransform()
71
+ if not transf.estimate(unit_square, vertices): raise Exception("estimate failed")
72
+ cropped_np_image = transform.warp(
73
+ np.array(example["image"]),
74
+ transf,
75
+ output_shape=(square_side + square_padding * 2, square_side + square_padding * 2)
76
+ )
77
+ cropped_image = Image.fromarray((cropped_np_image * 255).astype(np.uint8))
78
+ return {"cropped_image": cropped_image}
79
+
80
+ dataset = datasets.load_dataset("shortery/dm-codes")
81
+ dataset = dataset.map(crop_dm_code)
82
+ ```