jainr3 commited on
Commit
4b6c311
1 Parent(s): 4e600c9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +6 -29
README.md CHANGED
@@ -42,8 +42,7 @@ task_ids:
42
  - [Dataset Summary](#dataset-summary)
43
  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
44
  - [Languages](#languages)
45
- - [Two Subsets](#two-subsets)
46
- - [Key Differences](#key-differences)
47
  - [Dataset Structure](#dataset-structure)
48
  - [Data Instances](#data-instances)
49
  - [Data Fields](#data-fields)
@@ -52,13 +51,6 @@ task_ids:
52
  - [Data Splits](#data-splits)
53
  - [Loading Data Subsets](#loading-data-subsets)
54
  - [Method 1: Using Hugging Face Datasets Loader](#method-1-using-hugging-face-datasets-loader)
55
- - [Method 2. Use the PoloClub Downloader](#method-2-use-the-poloclub-downloader)
56
- - [Usage/Examples](#usageexamples)
57
- - [Downloading a single file](#downloading-a-single-file)
58
- - [Downloading a range of files](#downloading-a-range-of-files)
59
- - [Downloading to a specific directory](#downloading-to-a-specific-directory)
60
- - [Setting the files to unzip once they've been downloaded](#setting-the-files-to-unzip-once-theyve-been-downloaded)
61
- - [Method 3. Use `metadata.parquet` (Text Only)](#method-3-use-metadataparquet-text-only)
62
  - [Dataset Creation](#dataset-creation)
63
  - [Curation Rationale](#curation-rationale)
64
  - [Source Data](#source-data)
@@ -156,17 +148,13 @@ For example, below is the image of `f3501e05-aef7-4225-a9e9-f516527408ac.png` an
156
  ### Data Fields
157
 
158
  - key: Unique image name
159
- - `p`: Prompt
160
- - `se`: Random seed
161
- - `c`: CFG Scale (guidance scale)
162
- - `st`: Steps
163
- - `sa`: Sampler
164
 
165
  ### Dataset Metadata
166
 
167
  To help you easily access prompts and other attributes of images without downloading all the Zip files, we include a metadata table `metadata.parquet` for DiffusionDB-pixelart.
168
 
169
- The shape of `metadata.parquet` is (2000000, 13). Two tables share the same schema, and each row represents an image. We store these tables in the Parquet format because Parquet is column-based: you can efficiently query individual columns (e.g., prompts) without reading the entire table.
170
 
171
  Below are three random rows from `metadata.parquet`.
172
 
@@ -183,18 +171,7 @@ Below are three random rows from `metadata.parquet`.
183
  |Column|Type|Description|
184
  |:---|:---|:---|
185
  |`image_name`|`string`|Image UUID filename.|
186
- |`prompt`|`string`|The text prompt used to generate this image.|
187
- |`part_id`|`uint16`|Folder ID of this image.|
188
- |`seed`|`uint32`| Random seed used to generate this image.|
189
- |`step`|`uint16`| Step count (hyperparameter).|
190
- |`cfg`|`float32`| Guidance scale (hyperparameter).|
191
- |`sampler`|`uint8`| Sampler method (hyperparameter). Mapping: `{1: "ddim", 2: "plms", 3: "k_euler", 4: "k_euler_ancestral", 5: "k_heun", 6: "k_dpm_2", 7: "k_dpm_2_ancestral", 8: "k_lms", 9: "others"}`.
192
- |`width`|`uint16`|Image width.|
193
- |`height`|`uint16`|Image height.|
194
- |`user_name`|`string`|The unique discord ID's SHA256 hash of the user who generated this image. For example, the hash for `xiaohk#3146` is `e285b7ef63be99e9107cecd79b280bde602f17e0ca8363cb7a0889b67f0b5ed0`. "deleted_account" refer to users who have deleted their accounts. None means the image has been deleted before we scrape it for the second time.|
195
- |`timestamp`|`timestamp`|UTC Timestamp when this image was generated. None means the image has been deleted before we scrape it for the second time. Note that timestamp is not accurate for duplicate images that have the same prompt, hypareparameters, width, height.|
196
- |`image_nsfw`|`float32`|Likelihood of an image being NSFW. Scores are predicted by [LAION's state-of-art NSFW detector](https://github.com/LAION-AI/LAION-SAFETY) (range from 0 to 1). A score of 2.0 means the image has already been flagged as NSFW and blurred by Stable Diffusion.|
197
- |`prompt_nsfw`|`float32`|Likelihood of a prompt being NSFW. Scores are predicted by the library [Detoxicy](https://github.com/unitaryai/detoxify). Each score represents the maximum of `toxicity` and `sexual_explicit` (range from 0 to 1).|
198
 
199
  > **Warning**
200
  > Although the Stable Diffusion model has an NSFW filter that automatically blurs user-generated NSFW images, this NSFW filter is not perfect—DiffusionDB still contains some NSFW images. Therefore, we compute and provide the NSFW scores for images and prompts using the state-of-the-art models. The distribution of these scores is shown below. Please decide an appropriate NSFW score threshold to filter out NSFW images before using DiffusionDB in your projects.
@@ -217,8 +194,8 @@ You can use the Hugging Face [`Datasets`](https://huggingface.co/docs/datasets/q
217
  import numpy as np
218
  from datasets import load_dataset
219
 
220
- # Load the dataset with the `large_random_1k` subset
221
- dataset = load_dataset('jainr3/diffusiondb-pixelart', 'large_random_1k')
222
  ```
223
 
224
  ## Dataset Creation
 
42
  - [Dataset Summary](#dataset-summary)
43
  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
44
  - [Languages](#languages)
45
+ - [Subset](#subset)
 
46
  - [Dataset Structure](#dataset-structure)
47
  - [Data Instances](#data-instances)
48
  - [Data Fields](#data-fields)
 
51
  - [Data Splits](#data-splits)
52
  - [Loading Data Subsets](#loading-data-subsets)
53
  - [Method 1: Using Hugging Face Datasets Loader](#method-1-using-hugging-face-datasets-loader)
 
 
 
 
 
 
 
54
  - [Dataset Creation](#dataset-creation)
55
  - [Curation Rationale](#curation-rationale)
56
  - [Source Data](#source-data)
 
148
  ### Data Fields
149
 
150
  - key: Unique image name
151
+ - `p`: Text
 
 
 
 
152
 
153
  ### Dataset Metadata
154
 
155
  To help you easily access prompts and other attributes of images without downloading all the Zip files, we include a metadata table `metadata.parquet` for DiffusionDB-pixelart.
156
 
157
+ Two tables share the same schema, and each row represents an image. We store these tables in the Parquet format because Parquet is column-based: you can efficiently query individual columns (e.g., prompts) without reading the entire table.
158
 
159
  Below are three random rows from `metadata.parquet`.
160
 
 
171
  |Column|Type|Description|
172
  |:---|:---|:---|
173
  |`image_name`|`string`|Image UUID filename.|
174
+ |`text`|`string`|The text prompt used to generate this image.|
 
 
 
 
 
 
 
 
 
 
 
175
 
176
  > **Warning**
177
  > Although the Stable Diffusion model has an NSFW filter that automatically blurs user-generated NSFW images, this NSFW filter is not perfect—DiffusionDB still contains some NSFW images. Therefore, we compute and provide the NSFW scores for images and prompts using the state-of-the-art models. The distribution of these scores is shown below. Please decide an appropriate NSFW score threshold to filter out NSFW images before using DiffusionDB in your projects.
 
194
  import numpy as np
195
  from datasets import load_dataset
196
 
197
+ # Load the dataset with the `2k_random_1k` subset
198
+ dataset = load_dataset('jainr3/diffusiondb-pixelart', '2k_random_1k')
199
  ```
200
 
201
  ## Dataset Creation