Update README.md

README.md

@@ -11,18 +11,21 @@ tags:
 - fiftyone
 - image
 - image-segmentation
-dataset_summary: '
+- nasa
+- mars
+dataset_summary: >
 
 
 
 
-  This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 6000 samples.
+  This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 6000
+  samples.
 
 
   ## Installation
 
 
-  If you haven''t already, install FiftyOne:
+  If you haven't already, install FiftyOne:
 
 
   ```bash
@@ -44,9 +47,9 @@ dataset_summary: '
 
   # Load the dataset
 
-  # Note: other available arguments include ''max_samples'', etc
+  # Note: other available arguments include 'max_samples', etc
 
-  dataset = load_from_hub("harpreetsahota/S5Mars")
+  dataset = load_from_hub("Voxel51/S5Mars")
 
 
   # Launch the App
@@ -54,16 +57,11 @@ dataset_summary: '
   session = fo.launch_app(dataset)
 
   ```
-
-  '
 ---
 
-# Dataset Card for s5mars
-
-<!-- Provide a quick summary of the dataset. -->
-
-
+# Dataset Card for S5Mards Dataset
 
+![image/png](s5mars.gif)
 
 
 This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 6000 samples.
@@ -84,141 +82,228 @@ from fiftyone.utils.huggingface import load_from_hub
 
 # Load the dataset
 # Note: other available arguments include 'max_samples', etc
-dataset = load_from_hub("harpreetsahota/S5Mars")
+dataset = load_from_hub("Voxel51/S5Mars")
 
 # Launch the App
 session = fo.launch_app(dataset)
 ```
 
-
-## Dataset Details
-
 ### Dataset Description
 
-<!-- Provide a longer summary of what this dataset is. -->
+S5Mars (Semi-SuperviSed learning on Mars Semantic Segmentation) is a high-resolution Mars terrain semantic segmentation dataset containing 6,000 images captured by the Mars Science Laboratory (MSL) Curiosity rover. The dataset features sparse, high-confidence annotations designed to support semi-supervised learning approaches for Martian terrain analysis.
 
+The dataset addresses the challenge of limited high-quality annotations for Mars imagery by providing carefully curated, confidence-based sparse annotations. This enables research into semi-supervised learning methods that can leverage both labeled and unlabeled Mars terrain data for autonomous rover navigation and planning.
 
+Each image is annotated with pixel-level semantic segmentation masks covering 9 terrain classes commonly encountered in Martian landscapes.
 
-- **Curated by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Language(s) (NLP):** en
-- **License:** [More Information Needed]
+- **Curated by:** Jiahang Zhang, Lilang Lin, Zejia Fan, Wenjing Wang, Jiaying Liu (Peking University)
+- **License:** Research use
+- **Language(s):** N/A (Image dataset)
 
-### Dataset Sources [optional]
+### Dataset Sources
 
-<!-- Provide the basic links for the dataset. -->
-
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
+- **Repository:** [https://github.com/JHang2020/S5Mars_SSL](https://github.com/JHang2020/S5Mars_SSL)
+- **Paper:** [S5Mars: Semi-Supervised Learning for Mars Semantic Segmentation](https://ieeexplore.ieee.org/document/10499211) (IEEE TGRS 2024)
+- **Original Data Download:** [Google Drive](https://drive.google.com/file/d/130R5z9v2NChVuseNDsH0zSTMolteELkO/view)
 
 ## Uses
 
-<!-- Address questions around how the dataset is intended to be used. -->
-
 ### Direct Use
 
-<!-- This section describes suitable use cases for the dataset. -->
+This dataset is suitable for:
 
-[More Information Needed]
+- **Mars terrain semantic segmentation**: Training and evaluating models for pixel-level classification of Martian terrain
+- **Semi-supervised learning research**: Developing methods that leverage sparse annotations with unlabeled data
+- **Autonomous rover navigation**: Building perception systems for safe path planning on Mars
+- **Transfer learning**: Pre-training or fine-tuning models for planetary science applications
+- **Domain adaptation**: Studying domain shift between Earth and Mars imagery
 
 ### Out-of-Scope Use
 
-<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
-
-[More Information Needed]
+- Real-time mission-critical rover navigation without additional validation
+- Medical or safety-critical applications
+- Any use requiring absolute ground truth accuracy (annotations are sparse and confidence-based)
 
 ## Dataset Structure
 
-<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
+### Image Data
 
-[More Information Needed]
+- **Total images:** 6,000
+- **Resolution:** 1200 × 1200 pixels
+- **Format:** JPEG
+- **Source:** Mars Science Laboratory (MSL) Curiosity rover imagery
+- **Difficulty splits:** 
+  - `easy/`: 3,000 images with clearer terrain boundaries
+  - `hard/`: 3,000 images with more challenging segmentation scenarios
 
-## Dataset Creation
+### Annotations
 
-### Curation Rationale
+- **Format:** PNG semantic segmentation masks (1200 × 1200 pixels)
+- **Annotation style:** Sparse, confidence-based labeling
 
-<!-- Motivation for the creation of this dataset. -->
+### Semantic Classes
 
-[More Information Needed]
+| Pixel Value | Class Name | Description |
+|-------------|------------|-------------|
+| 0 | background | Unlabeled/null regions |
+| 1 | sky | Martian sky |
+| 2 | ridge | Ridge formations |
+| 3 | soil | Martian soil |
+| 4 | sand | Sandy terrain |
+| 5 | bedrock | Exposed bedrock |
+| 6 | rock | Individual rocks |
+| 7 | rover | Rover components visible in frame |
+| 8 | trace | Rover wheel tracks/traces |
+| 9 | hole | Holes or depressions |
 
-### Source Data
+### Data Splits
 
-<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
+| Split | Samples | Description |
+|-------|---------|-------------|
+| Train | 5,000 | Training set |
+| Validation | 500 | Validation set |
+| Test | 500 | Held-out test set |
 
-#### Data Collection and Processing
+### FiftyOne Dataset Format
 
-<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
+This dataset is provided in [FiftyOne](https://docs.voxel51.com/) format for easy visualization, exploration, and integration with ML workflows.
 
-[More Information Needed]
+#### Sample Fields
 
-#### Who are the source data producers?
+| Field | Type | Description |
+|-------|------|-------------|
+| `filepath` | `string` | Absolute path to the image file |
+| `tags` | `list[string]` | Data split membership: `["train"]`, `["val"]`, or `["test"]` |
+| `difficulty` | `string` | Image difficulty category: `"easy"` or `"hard"` |
+| `ground_truth` | `fo.Segmentation` | Semantic segmentation mask |
+| `metadata` | `fo.ImageMetadata` | Image dimensions, size, etc. |
 
-<!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->
+#### Mask Targets
 
-[More Information Needed]
+The `ground_truth` segmentation field uses the following `mask_targets` mapping:
 
-### Annotations [optional]
+```python
+dataset.mask_targets = {
+    "ground_truth": {
+        1: "sky",
+        2: "ridge",
+        3: "soil",
+        4: "sand",
+        5: "bedrock",
+        6: "rock",
+        7: "rover",
+        8: "trace",
+        9: "hole",
+    }
+}
+```
 
-<!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
+**Note:** Pixel value `0` represents background/unlabeled regions and is rendered as invisible in FiftyOne's App.
 
-#### Annotation process
+#### Loading the Dataset
 
-<!-- This section describes the annotation process such as annotation tools used in the process, the amount of data annotated, annotation guidelines provided to the annotators, interannotator statistics, annotation validation, etc. -->
+```python
+import fiftyone as fo
 
-[More Information Needed]
+# Load the dataset
+dataset = fo.load_dataset("s5mars")
 
-#### Who are the annotators?
+# View dataset info
+print(dataset)
 
-<!-- This section describes the people or systems who created the annotations. -->
+# Filter by split
+train_view = dataset.match_tags("train")
+test_view = dataset.match_tags("test")
 
-[More Information Needed]
+# Filter by difficulty
+easy_view = dataset.match(F("difficulty") == "easy")
+hard_view = dataset.match(F("difficulty") == "hard")
 
-#### Personal and Sensitive Information
+# Launch the App
+session = fo.launch_app(dataset)
+```
 
-<!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. -->
+## Dataset Creation
 
-[More Information Needed]
+### Curation Rationale
 
-## Bias, Risks, and Limitations
+Deep learning has become a powerful tool for Mars exploration, with terrain semantic segmentation being fundamental for rover autonomous planning and safe driving. However, there is a lack of sufficient detailed and high-confidence data annotations required by most deep learning methods. 
 
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+S5Mars addresses this gap by providing:
+1. High-resolution imagery from actual Mars rover missions
+2. Sparse but high-confidence annotations based on annotator certainty
+3. A benchmark for semi-supervised learning approaches tailored to Mars imagery
 
-[More Information Needed]
+### Source Data
 
-### Recommendations
+#### Data Collection and Processing
 
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+Images were collected from the Mars Science Laboratory (MSL) Curiosity rover's navigation and science cameras. The dataset includes both "easy" and "hard" subsets, categorized based on the complexity of terrain boundaries and segmentation difficulty.
 
-Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations.
+#### Who are the source data producers?
 
-## Citation [optional]
+- **Original imagery:** NASA/JPL-Caltech Mars Science Laboratory mission
+- **Dataset curation:** Researchers at Peking University
 
-<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
+### Annotations
 
-**BibTeX:**
+#### Annotation process
 
-[More Information Needed]
+Annotations follow a sparse, confidence-based labeling strategy:
+- Annotators label regions where they have high confidence in the terrain classification
+- Ambiguous or uncertain regions are left unlabeled (pixel value 0)
+- This approach ensures high label quality at the cost of completeness
 
-**APA:**
+#### Who are the annotators?
 
-[More Information Needed]
+Annotations were created by the research team at Peking University as part of the S5Mars project.
 
-## Glossary [optional]
+#### Personal and Sensitive Information
+
+This dataset contains no personal or sensitive information. All images are of Martian terrain captured by NASA's publicly available rover imagery.
+
+## Bias, Risks, and Limitations
+
+- **Sparse annotations:** Not all pixels are labeled; unlabeled regions should not be treated as negative examples
+- **Class imbalance:** Sky and common terrain types dominate; rare classes like "rover" and "hole" have fewer samples
+- **Sensor-specific:** Images are from MSL Curiosity rover cameras; may not generalize to other Mars missions
+- **Lighting conditions:** Martian lighting varies; model performance may vary across different times of day
+- **Annotation subjectivity:** Terrain class boundaries can be ambiguous (e.g., soil vs. sand)
+
+### Recommendations
+
+- Use semi-supervised learning approaches to leverage unlabeled regions
+- Consider class-weighted losses to handle imbalance
+- Validate on the held-out test set before deployment
+- For critical applications, combine with additional human review
+
+## Citation
 
-<!-- If relevant, include terms and calculations in this section that can help readers understand the dataset or dataset card. -->
+**BibTeX:**
 
-[More Information Needed]
+```bibtex
+@ARTICLE{10499211,
+  author={Zhang, Jiahang and Lin, Lilang and Fan, Zejia and Wang, Wenjing and Liu, Jiaying},
+  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
+  title={S5Mars: Semi-Supervised Learning for Mars Semantic Segmentation}, 
+  year={2024},
+  volume={62},
+  pages={1-15},
+  doi={10.1109/TGRS.2024.33870211}
+}
+```
 
-## More Information [optional]
+**APA:**
 
-[More Information Needed]
+Zhang, J., Lin, L., Fan, Z., Wang, W., & Liu, J. (2024). S5Mars: Semi-Supervised Learning for Mars Semantic Segmentation. *IEEE Transactions on Geoscience and Remote Sensing*, 62, 1-15.
 
-## Dataset Card Authors [optional]
+## Glossary
 
-[More Information Needed]
+- **MSL:** Mars Science Laboratory, NASA's Mars rover mission featuring the Curiosity rover
+- **Semi-supervised learning:** Machine learning approach using both labeled and unlabeled data
+- **Semantic segmentation:** Pixel-level classification task assigning a class label to each pixel
+- **Sparse annotation:** Labeling strategy where only confident regions are annotated
 
 ## Dataset Card Contact
 
-[More Information Needed]
+For questions about the original dataset, please refer to the [GitHub repository](https://github.com/JHang2020/S5Mars_SSL) or contact the paper authors.