Datasets:
timm
/

Modalities:
Image
Formats:
parquet
ArXiv:
Libraries:
Datasets
Dask
License:
rwightman HF staff commited on
Commit
bd8779f
1 Parent(s): c428563

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +15 -5
README.md CHANGED
@@ -109,16 +109,16 @@ dataset_info:
109
  '99': n13133613
110
  splits:
111
  - name: train
112
- num_bytes: 6284840508.0
113
  num_examples: 50000
114
  - name: validation
115
- num_bytes: 1286953696.0
116
  num_examples: 10000
117
  - name: test
118
- num_bytes: 670707560.0
119
  num_examples: 5000
120
  download_size: 7433461683
121
- dataset_size: 8242501764.0
122
  configs:
123
  - config_name: default
124
  data_files:
@@ -134,6 +134,8 @@ license_link: https://www.image-net.org/download.php
134
  task_categories:
135
  - image-classification
136
  pretty_name: Mini-ImageNet
 
 
137
  ---
138
 
139
  ## Dataset Description
@@ -153,9 +155,17 @@ Unlike some 'mini' variants this one includes the original images at their origi
153
  #### Test
154
  * 5000 samples from ImageNet-1k validation split (all 50 samples per class)
155
 
 
 
 
 
 
 
 
 
156
  ### Citation Information
157
 
158
- For the specific instance of this mini variant I am not sure what the origin is. It is different from commonly referred [Vinyales et al., 2016](https://arxiv.org/abs/1606.04080) as it doesn't match the classes / splits.
159
 
160
  Train & validation splits match train & test of https://www.kaggle.com/datasets/ctrnngtrung/miniimagenet ... it is not clear where that originated though.
161
 
 
109
  '99': n13133613
110
  splits:
111
  - name: train
112
+ num_bytes: 6284840508
113
  num_examples: 50000
114
  - name: validation
115
+ num_bytes: 1286953696
116
  num_examples: 10000
117
  - name: test
118
+ num_bytes: 670707560
119
  num_examples: 5000
120
  download_size: 7433461683
121
+ dataset_size: 8242501764
122
  configs:
123
  - config_name: default
124
  data_files:
 
134
  task_categories:
135
  - image-classification
136
  pretty_name: Mini-ImageNet
137
+ size_categories:
138
+ - 10K<n<100K
139
  ---
140
 
141
  ## Dataset Description
 
155
  #### Test
156
  * 5000 samples from ImageNet-1k validation split (all 50 samples per class)
157
 
158
+ ### Usage
159
+
160
+ This dataset is good for testing hparams and models in `timm`
161
+
162
+ #### Train
163
+
164
+ `python train.py --dataset hfds/timm/mini-imagenet --model resnet50 --amp --num-classes 100`
165
+
166
  ### Citation Information
167
 
168
+ For the specific instance of this mini variant I am not sure what the origin is. It is different from commonly referenced [Vinyales et al.,2016](https://arxiv.org/abs/1606.04080) as it doesn't match the classes / splits.
169
 
170
  Train & validation splits match train & test of https://www.kaggle.com/datasets/ctrnngtrung/miniimagenet ... it is not clear where that originated though.
171