javyduck
commited on
Commit
β’
494dcce
1
Parent(s):
cab5e0c
update
Browse files- .ipynb_checkpoints/README-checkpoint.md +45 -0
- README.md +45 -3
.ipynb_checkpoints/README-checkpoint.md
ADDED
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# MMCBench Dataset: Benchmarking Dataset for Multimodal Model Evaluation π
|
2 |
+
|
3 |
+
## Overview
|
4 |
+
|
5 |
+
The MMCBench Dataset is a curated collection of data designed for the comprehensive evaluation of Large Multimodal Models (LMMs) under common corruption scenarios. This dataset supports the MMCBench framework, focusing on cross-modal interactions involving text, image, and speech. It provides essential data for generative tasks such as text-to-image, image-to-text, text-to-speech, and speech-to-text, enabling robustness and self-consistency assessments of LMMs.
|
6 |
+
|
7 |
+
## Dataset Composition π
|
8 |
+
|
9 |
+
The MMCBench Dataset is structured to facilitate the evaluation across four key generative tasks:
|
10 |
+
|
11 |
+
- **Text-to-Image:** A collection of text descriptions with their corresponding corrupted versions and associated images.
|
12 |
+
- **Image-to-Text:** A set of images with clean and corrupted captions.
|
13 |
+
- **Text-to-Speech:** Text inputs with their clean and corrupted audio outputs.
|
14 |
+
- **Speech-to-Text:** Audio files with transcriptions before and after audio corruptions.
|
15 |
+
|
16 |
+
Each subset of the dataset has been meticulously selected and processed to represent challenging scenarios for LMMs.
|
17 |
+
|
18 |
+
## Using the Dataset π οΈ
|
19 |
+
|
20 |
+
To use the MMCBench Dataset for model evaluation:
|
21 |
+
|
22 |
+
1. **Access the Data**: The dataset is hosted on Hugging Face and can be accessed using their dataset library or direct download.
|
23 |
+
2. **Select the Task**: Choose from text-to-image, image-to-text, text-to-speech, or speech-to-text tasks based on your model's capabilities.
|
24 |
+
3. **Apply the Benchmark**: Utilize the data for each task to test your model's performance against various corruptions. Follow the [MMCBench](https://github.com/sail-sg/MMCBench/tree/main) framework for a consistent and standardized evaluation.
|
25 |
+
|
26 |
+
### Dataset Structure π
|
27 |
+
|
28 |
+
The dataset is organized into four main directories, each corresponding to one of the generative tasks:
|
29 |
+
|
30 |
+
- `text2image/`: Contains text inputs and associated images.
|
31 |
+
- `image2text/`: Comprises images and their descriptive captions.
|
32 |
+
- `text2speech/`: Includes text inputs and generated speech outputs.
|
33 |
+
- `speech2text/`: Contains audio files and their transcriptions.
|
34 |
+
|
35 |
+
## Contributing to the Dataset π€
|
36 |
+
|
37 |
+
Contributions to the MMCBench Dataset are welcome. If you have suggestions for additional data or improvements, please reach out through the Hugging Face platform or directly contribute via GitHub.
|
38 |
+
|
39 |
+
## License π
|
40 |
+
|
41 |
+
The MMCBench Dataset is made available under the Apache 2.0 License, ensuring open and ethical use for research and development.
|
42 |
+
|
43 |
+
## Acknowledgments and Citations π
|
44 |
+
|
45 |
+
When using the MMCBench Dataset in your research, please cite it appropriately. We extend our gratitude to all contributors and collaborators who have enriched this dataset, making it a valuable resource for the AI and ML community.
|
README.md
CHANGED
@@ -1,3 +1,45 @@
|
|
1 |
-
|
2 |
-
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# MMCBench Dataset: Benchmarking Dataset for Multimodal Model Evaluation π
|
2 |
+
|
3 |
+
## Overview
|
4 |
+
|
5 |
+
The MMCBench Dataset is a curated collection of data designed for the comprehensive evaluation of Large Multimodal Models (LMMs) under common corruption scenarios. This dataset supports the MMCBench framework, focusing on cross-modal interactions involving text, image, and speech. It provides essential data for generative tasks such as text-to-image, image-to-text, text-to-speech, and speech-to-text, enabling robustness and self-consistency assessments of LMMs.
|
6 |
+
|
7 |
+
## Dataset Composition π
|
8 |
+
|
9 |
+
The MMCBench Dataset is structured to facilitate the evaluation across four key generative tasks:
|
10 |
+
|
11 |
+
- **Text-to-Image:** A collection of text descriptions with their corresponding corrupted versions and associated images.
|
12 |
+
- **Image-to-Text:** A set of images with clean and corrupted captions.
|
13 |
+
- **Text-to-Speech:** Text inputs with their clean and corrupted audio outputs.
|
14 |
+
- **Speech-to-Text:** Audio files with transcriptions before and after audio corruptions.
|
15 |
+
|
16 |
+
Each subset of the dataset has been meticulously selected and processed to represent challenging scenarios for LMMs.
|
17 |
+
|
18 |
+
## Using the Dataset π οΈ
|
19 |
+
|
20 |
+
To use the MMCBench Dataset for model evaluation:
|
21 |
+
|
22 |
+
1. **Access the Data**: The dataset is hosted on Hugging Face and can be accessed using their dataset library or direct download.
|
23 |
+
2. **Select the Task**: Choose from text-to-image, image-to-text, text-to-speech, or speech-to-text tasks based on your model's capabilities.
|
24 |
+
3. **Apply the Benchmark**: Utilize the data for each task to test your model's performance against various corruptions. Follow the [MMCBench](https://github.com/sail-sg/MMCBench/tree/main) framework for a consistent and standardized evaluation.
|
25 |
+
|
26 |
+
### Dataset Structure π
|
27 |
+
|
28 |
+
The dataset is organized into four main directories, each corresponding to one of the generative tasks:
|
29 |
+
|
30 |
+
- `text2image/`: Contains text inputs and associated images.
|
31 |
+
- `image2text/`: Comprises images and their descriptive captions.
|
32 |
+
- `text2speech/`: Includes text inputs and generated speech outputs.
|
33 |
+
- `speech2text/`: Contains audio files and their transcriptions.
|
34 |
+
|
35 |
+
## Contributing to the Dataset π€
|
36 |
+
|
37 |
+
Contributions to the MMCBench Dataset are welcome. If you have suggestions for additional data or improvements, please reach out through the Hugging Face platform or directly contribute via GitHub.
|
38 |
+
|
39 |
+
## License π
|
40 |
+
|
41 |
+
The MMCBench Dataset is made available under the Apache 2.0 License, ensuring open and ethical use for research and development.
|
42 |
+
|
43 |
+
## Acknowledgments and Citations π
|
44 |
+
|
45 |
+
When using the MMCBench Dataset in your research, please cite it appropriately. We extend our gratitude to all contributors and collaborators who have enriched this dataset, making it a valuable resource for the AI and ML community.
|