Update README.md

README.md

@@ -34,9 +34,9 @@ description: >-
 
 # Dataset Card for GZCD (Grevy's Zebra Census Dataset)
 
-The **GZCD (Grevy's Zebra Census Dataset)** is a comprehensive, multi-modal ecological computer vision dataset designed for wildlife detection, species classification, and individual animal re-identification (Re-ID). Captured primarily in the field, this dataset emphasizes endangered equid and ungulate populations, specifically Grevy's zebras, Plains zebras, and Reticulated giraffes. 
+The **GZCD (Grevy's Zebra Census Dataset)** is a comprehensive ecological computer vision dataset designed for wildlife detection, species classification, and individual animal re-identification (Re-ID). Captured primarily in the field, this dataset emphasizes endangered equid and ungulate populations, specifically Grevy's zebras, Plains zebras, and Reticulated giraffes. 
 
-GZCD consists of field photographs taken during the 2016 and 2018 iterations of the **Great Grevy's Rally (GGR)**. The data collection was carried out by citizen scientists and conservationists across a 25,000 square kilometer range. It was curated to help test and train new computer vision algorithms, most notably building upon the pivotal animal re-identification work by Jason Parham.
+GZCD consists of field photographs captured in Meru County, Kenya, taken over four days by 13 photographers during 2016 and 2018 iterations of the **Great Grevy's Rally (GGR)**. The data collection was carried out by citizen scientists and conservationists across a 25,000 square kilometer range. It was curated to help train and test new computer vision algorithms, most notably building upon the pivotal animal re-identification work by Jason Parham.
 
 ---
 
@@ -47,7 +47,7 @@ GZCD consists of field photographs taken during the 2016 and 2018 iterations of
 - **Repository:** [Imageomics/GZCD](https://huggingface.co/datasets/imageomics/GZCD)
 - **Pipeline Code:** [Imageomics/GGR_pipeline](https://github.com/Imageomics/GGR_pipeline/tree/main)
 
-This dataset provides raw ecological image data alongside the deeply processed outputs of the **GGR Pipeline**. It serves as an ideal resource for researchers bridging Computer Vision (CV) and ecology, providing ground-truth and machine-generated annotations for bounding boxes, viewpoints, identifiability scores, and individual animal Re-ID clusters.
+This dataset provides raw ecological image data alongside the carefully processed outputs of the **GGR Pipeline**. It serves as an ideal resource for researchers bridging Computer Vision (CV) and ecology, providing manual ground-truth and annotations for bounding boxes, viewpoints, census annotation scores, and individual animal Re-ID clusters obtained after applying the semi-automated GGR pipeline.
 
 ---
 
@@ -70,7 +70,7 @@ text
 ### Data Instances
 - `images/`: The raw field images capturing various wildlife species.
 - `initial_metadata.json`: Represents the raw state of the dataset prior to processing, containing EXIF data, camera body details, flash card source data, GPS coordinates, and initial image dimensions.
-- `final_metadata.json`: The fully annotated dataset resulting from the 9-stage CV pipeline, containing localized bounding boxes, classifications, and individual tracking IDs.
+- `final_metadata.json`: The fully annotated dataset resulting from the 9-stage CV pipeline, containing localized bounding boxes, species, viewpoint and census annotation classifications, and individual IDs.
 
 ### Data Fields
 
@@ -142,20 +142,18 @@ The images were collected in **Meru County, Kenya** during the 2016 and 2018 cen
 2. **Detection:** Utilizes YOLO to localize animals in the images, creating initial bounding boxes (`bbox`).
 3. **Species Classification:** Processes detections through BioCLIP to classify the species (e.g., Grevy's zebra vs. Plains zebra).
 4. **Viewpoint Classification:** Determines the orientation of the detected animal (`up`, `front`, `back`, `left`, `right`).
-5. **Identifiable Annotation (IA) Classification:** Evaluates photographic quality and pose. Assigns a `CA_score` and a boolean flag indicating if the animal's features (e.g., stripes) are clear enough for individual identification.
-7. **IA Filtering:** Discards annotations below the identifiability threshold and simplifies viewpoints strictly to `left` or `right` flanks for Re-ID purposes.
-8. **Miew-ID:** Generates deep metric embeddings for all remaining high-quality annotations leveraging the Miew-Id feature extraction model.
-9. **Local Clusters and Alternatives (LCA) Algorithm:** Clusters the generated embeddings based on feature similarity, grouping images of the same individual animal together.
-10. **Post-processing and ID Assignment:** Applies biological consistency checks, resolves cluster overlaps, allows for human-in-the-loop manual verification, assigns the final `individual_id`, and integrates non-identifiable tracking links where necessary.
-
+5. **Census Annotation (CA) Classification:** Evaluates photographic quality and pose. Assigns a `CA_score` and a boolean flag indicating if the animal's features (e.g., stripes) are clear enough for individual identification.
+6. **CA Filtering:** Discards annotations below the identifiability threshold decided empirically and simplifies viewpoints strictly to `left` or `right` flanks for Re-ID purposes.
+7. **MiewID:** Generates deep metric embeddings for all filtered high-quality annotations leveraging the MiewID feature extraction model.
+8. **Local Clusters and Alternatives (LCA) Algorithm:** Clusters the generated embeddings based on feature similarity, grouping annotations of the same individual animal together by resolving inconsistencies.
 ---
 
 ## Considerations for Using the Data
 
 - **Missing Location Data:** When GPS coordinates could not be resolved from field equipment or manual logs, fields for `latitude` and `longitude` (`gps_lat`/`gps_lon`) default to `-1.0`. Algorithms utilizing geometric filtering should catch these sentinel values.
-- **Viewpoint Filtering:** Because animal Re-ID relies heavily on flank stripe patterns, the pipeline aggressively filters for `left` and `right` viewpoints. Forward or backward-facing animals are detected but may be excluded from Re-ID tracking.
-- **Geographic Focus:** The spatial data anchors tightly to the **Meru County** ecosystem and adjacent northern rangelands in Kenya. Models trained heavily on this dataset may experience domain shift when deployed in different biomes or under disparate lighting conditions.
-- **Class Imbalance:** Due to the targeted focus of the Great Grevy's Rally, equid and giraffe classes heavily outnumber categories like `lion` or `rhino_black`.
+- **Viewpoint Filtering:** Because Re-ID of Grevy's Zebra relies on right-flank stripe patterns, the pipeline aggressively selects `right` viewpoints whose embeddings are extracted and ID assigned. Other viewpoints are excluded from ID assignment.
+- **Geographic Focus:** The spatial data anchors tightly to the **Meru County** ecosystem in Kenya. Models trained heavily on this dataset may experience domain shift when deployed in different biomes or under disparate lighting conditions.
+- **Class Imbalance:** Due to the targeted focus of the Great Grevy's Rally on Grevy's Zebras, the final JSON contains individual IDs for Grevy's Zebra only.
 
 ---