sps44 commited on
Commit
a68df04
1 Parent(s): de3dd6c

Readme update with DCAI intro

Browse files
Files changed (2) hide show
  1. README.md +53 -13
  2. data/spotlight_save_layout.png +3 -0
README.md CHANGED
@@ -58,33 +58,73 @@ tags:
58
 
59
  ### Dataset Summary
60
 
61
- This dataset ...
62
 
63
- [//]: # (todo: verantwortlichkeit/attribution)
 
 
 
64
 
65
- ### Explore DCASE23 Task 2 with Spotlight
 
 
 
 
 
 
66
 
67
  ![Analyze DCASE23 Task 2 with Spotlight](data/preview_dcase.png "Analyze DCASE23 Task 2 with Spotlight")
68
 
69
- Install Spotlight via [pip](https://packaging.python.org/en/latest/key_projects/#pip)
70
- ```bash
71
- pip install renumics-spotlight
 
 
 
 
72
  ```
73
 
74
- and simply run the following codeblock:
 
 
 
 
 
75
  ```jupyterpython
76
- from datasets import load_dataset, load_dataset_builder
77
  from renumics import spotlight
78
 
79
- dataset = load_dataset("renumics/dcase23-task2-enriched", "dev", split="all", streaming=False)
80
- ds_builder = load_dataset_builder("renumics/dcase23-task2-enriched", "dev")
 
 
 
 
 
81
 
82
- df = ds_builder.config.to_spotlight(dataset)
 
 
 
83
  spotlight.show(df, dtype={'audio': spotlight.Audio, "ast-finetuned-audioset-10-10-0.4593-embeddings": spotlight.Embedding}, layout=ds_builder.config.get_layout(config="standard"))
84
  ```
85
 
86
- [//]: # (todo: how2 -> dataset -> spotlight)
87
- [//]: # (todo: spotlight)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
88
 
89
  ## Dataset Structure
90
 
 
58
 
59
  ### Dataset Summary
60
 
61
+ [Data-centric AI](https://datacentricai.org) principles have become increasingly important for real-world use cases. At [Renumics](https://renumics.com/) we believe that classical benchmark datasets and competitions should be extended to reflect this development.
62
 
63
+ This is why we are publishing benchmark datasets with application-specific enrichments (e.g. embeddings, baseline results, uncertainties, label error scores). We hope this helps the ML community in the following ways:
64
+ 1. Enable new researchers to quickly develop a profound understanding of the dataset.
65
+ 2. Popularize data-centric AI principles and tooling in the ML community.
66
+ 3. Encourage the sharing of meaningful qualitative insights in addition to traditional quantitative metrics.
67
 
68
+ This dataset is an enriched version of the [dataset](https://zenodo.org/record/7690148#.ZAXsSdLMLmE) provided in the context of the [anomalous sound detection task](https://dcase.community/challenge2023/task-first-shot-unsupervised-anomalous-sound-detection-for-machine-condition-monitoring) of the [DCASE2023 challenge](https://dcase.community/challenge2023/). The enrichment include an embedding generated by a pre-trained [Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer#transformers.ASTFeatureExtractor) and results of the official challenge [baseline implementation](https://github.com/nttcslab/dase2023_task2_baseline_ae).
69
+
70
+ ### DCASE23 Task2 Dataset
71
+
72
+ Once a year, the [DCASE community](https://dcase.community/) publishes a [challenge](https://dcase.community/challenge2023/) with several tasks in the context of acoustic event detection and classification. [Task 2 of this challenge](https://dcase.community/challenge2023/task-first-shot-unsupervised-anomalous-sound-detection-for-machine-condition-monitoring) deals with anomalous sound detection for machine condition monitoring. The original dataset is based on the [MIMII DG](https://arxiv.org/abs/2205.13879) and the [ToyADMOS2](https://arxiv.org/abs/2106.02369) datasets. Please cite the papers by [Harada et al.](https://arxiv.org/abs/2106.02369) and [Dohi et al.](https://arxiv.org/abs/2205.13879) if you use this dataset and the paper by [Harada et al.](https://arxiv.org/pdf/2303.00455.pdf) if you use the baseline results.
73
+
74
+ ### Explore Dataset
75
 
76
  ![Analyze DCASE23 Task 2 with Spotlight](data/preview_dcase.png "Analyze DCASE23 Task 2 with Spotlight")
77
 
78
+ The enrichments allow you to quickly gain insights into the dataset. The open source data curation tool Renumics Spotlight enables that with just a few lines of code:
79
+
80
+ Load the dataset from huggingface in your notebook:
81
+ ```jupyterpython
82
+ import datasets
83
+
84
+ dataset = datasets.load_dataset("renumics/dcase23-task2-enriched", "dev", split="all", streaming=False)
85
  ```
86
 
87
+ Install Spotlight via [pip](https://packaging.python.org/en/latest/key_projects/#pip):
88
+ ```jupyterpython
89
+ !pip install renumics-spotlight
90
+ ```
91
+
92
+ Start exploring with a simple view that leverages embeddings to identify relevant data segments:
93
  ```jupyterpython
 
94
  from renumics import spotlight
95
 
96
+ df = dataset.to_pandas()
97
+ spotlight.show(df, dtype={'audio': spotlight.Audio, "ast-finetuned-audioset-10-10-0.4593-embeddings": spotlight.Embedding}, layout=ds_builder.config.get_layout(config="standard"))
98
+ ```
99
+ You can use the UI to interactively configure the view on the data. Depending on the concrete taks (e.g. model comparison, debugging, outlier detection) you might want to leverage different enrichments and metadata.
100
+
101
+ This example shows a view that is suitable to compare the results of two different ML models:
102
+
103
 
104
+ ```jupyterpython
105
+ from renumics import spotlight
106
+
107
+ df = dataset.to_pandas()
108
  spotlight.show(df, dtype={'audio': spotlight.Audio, "ast-finetuned-audioset-10-10-0.4593-embeddings": spotlight.Embedding}, layout=ds_builder.config.get_layout(config="standard"))
109
  ```
110
 
111
+ ## Using custom model results and enrichments
112
+
113
+ When developing your custom model you want to use different kinds of information from you model (e.g. embedding, anomaly scores etc.) to gain further insights into the dataset and the model behvior.
114
+
115
+ Suppose you have your model's embeddings for each datapoint as a 2D-Numpy array and your anomaly score as a 1D-Numpy array. Then you can add this information to the dataset:
116
+ ```jupyterpython
117
+ from renumics import spotlight
118
+ #todo
119
+ ```
120
+ Depending on your concrete task you might want to use different enrichments. For a good overview on great open source tooling for uncertainty quantification, explainability and outlier detection, you can take a look at our [curated list for open source data-centric AI tooling](https://github.com/Renumics/awesome-open-data-centric-ai) on Github.
121
+
122
+ You can also save your view configuration in Spotlight in a JSON configuration file by clicking on the respective icon:
123
+
124
+ ![Save a data curation layout in Spotlight](data/spotlight_save_layout.png "Save a data curation layout in Spotlight")
125
+
126
+ We have also set up a Github repo with interactive notebooks that highlight different enrichment possibilities and curation templates. We encourage you to submit a pull request to share your enrichment code and interaction layouts.
127
+
128
 
129
  ## Dataset Structure
130
 
data/spotlight_save_layout.png ADDED

Git LFS Details

  • SHA256: 99bdaa1df1c2c8f2cb648920336ceb198ccdfa6d97b5078bcf2dac556dad04ca
  • Pointer size: 130 Bytes
  • Size of remote file: 14.7 kB