tanganke commited on
Commit
b367b8b
1 Parent(s): f6c366d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +50 -0
README.md CHANGED
@@ -56,4 +56,54 @@ configs:
56
  path: data/train-*
57
  - split: val
58
  path: data/val-*
 
 
 
 
 
 
59
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
56
  path: data/train-*
57
  - split: val
58
  path: data/val-*
59
+ task_categories:
60
+ - depth-estimation
61
+ - image-segmentation
62
+ - image-feature-extraction
63
+ size_categories:
64
+ - 1K<n<10K
65
  ---
66
+
67
+
68
+ This is the NYUv2 dataset for scene understanding tasks.
69
+ I downloaded the original data from the [Tsinghua Cloud](https://cloud.tsinghua.edu.cn/f/6d0a89f4ca1347d8af5f/?dl=1) and transformed it into Huggingface Dataset.
70
+ Credit to [ForkMerge: Mitigating Negative Transfer in Auxiliary-Task Learning](http://arxiv.org/abs/2301.12618).
71
+
72
+ ## Dataset Information
73
+
74
+ This data contains two splits: 'train' and 'val' (used as test dataset).
75
+ Each sample in the dataset has 5 items: 'image', 'segmentation', 'depth', 'normal', and 'noise'.
76
+ The noise is generated using `torch.rand()`.
77
+
78
+
79
+ ## Usage
80
+
81
+ ```python
82
+ dataset = load_dataset('tanganke/nyuv2')
83
+ dataset = dataset.with_format('torch') # this will convert the items into `torch.Tensor` objects
84
+ ```
85
+
86
+ this will return a `DatasetDict`:
87
+
88
+ ```python
89
+ DatasetDict({
90
+ train: Dataset({
91
+ features: ['image', 'segmentation', 'depth', 'normal', 'noise'],
92
+ num_rows: 795
93
+ })
94
+ val: Dataset({
95
+ features: ['image', 'segmentation', 'depth', 'normal', 'noise'],
96
+ num_rows: 654
97
+ })
98
+ })
99
+ ```
100
+
101
+ The features:
102
+
103
+ ```python
104
+ {'image': Array3D(shape=(3, 288, 384), dtype='float32', id=None),
105
+ 'segmentation': Array2D(shape=(288, 384), dtype='int64', id=None),
106
+ 'depth': Array3D(shape=(1, 288, 384), dtype='float32', id=None),
107
+ 'normal': Array3D(shape=(3, 288, 384), dtype='float32', id=None),
108
+ 'noise': Array3D(shape=(1, 288, 384), dtype='float32', id=None)}
109
+ ```