Update README.md

README.md

@@ -46,7 +46,7 @@ dataset_summary: '
 
   # Note: other available arguments include ''max_samples'', etc
 
-  dataset = load_from_hub("harpreetsahota/BarkVN-50")
+  dataset = load_from_hub("Voxel51/BarkVN-50")
 
 
   # Launch the App
@@ -58,12 +58,9 @@ dataset_summary: '
   '
 ---
 
-# Dataset Card for BarkVN-50
-
-<!-- Provide a quick summary of the dataset. -->
-
-
+# Dataset Card for BarkVN-50: Tree Species Identification from Bark Texture
 
+![image/png](barkvn50.gif)
 
 
 This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 5578 samples.
@@ -84,7 +81,7 @@ from fiftyone.utils.huggingface import load_from_hub
 
 # Load the dataset
 # Note: other available arguments include 'max_samples', etc
-dataset = load_from_hub("harpreetsahota/BarkVN-50")
+dataset = load_from_hub("Voxel51/BarkVN-50")
 
 # Launch the App
 session = fo.launch_app(dataset)
@@ -95,130 +92,149 @@ session = fo.launch_app(dataset)
 
 ### Dataset Description
 
-<!-- Provide a longer summary of what this dataset is. -->
+BarkVN-50 is a comprehensive tree bark texture classification dataset containing 5,678 images of 303×404 pixels across 50 different tree species. This dataset represents the maximum number of tree species considered for bark classification research at the time of its creation. The dataset was specifically designed for automated tree species identification using computer vision and deep learning techniques.
 
+Bark texture provides a reliable characteristic for tree identification as it remains present despite seasonal variations, unlike leaves, fruits, or flowers. This makes the dataset particularly valuable for year-round forestry applications including forest conservation, disease diagnosis, and plant production.
 
-
-- **Curated by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Language(s) (NLP):** en
+- **Curated by:** Truong Hoang, Vinh
+- **Original Dataset:** [BarkVN-50 on Mendeley Data](https://data.mendeley.com/datasets/gbt4tdmttn/1)
+- **Paper:** [Automated Identification of Tree Species by Bark Texture Classification Using Convolutional Neural Networks](https://arxiv.org/abs/2210.09290v1) by Sahil Faizal
 - **License:** [More Information Needed]
 
-### Dataset Sources [optional]
-
-<!-- Provide the basic links for the dataset. -->
+### Dataset Sources
 
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
-
-## Uses
-
-<!-- Address questions around how the dataset is intended to be used. -->
+- **Repository:** https://data.mendeley.com/datasets/gbt4tdmttn/1
+- **Paper:** https://arxiv.org/abs/2210.09290v1
+- **DOI:** 10.17632/gbt4tdmttn.1
+## Dataset Structure
 
-### Direct Use
+### Original Dataset
 
-<!-- This section describes suitable use cases for the dataset. -->
+The original BarkVN-50 dataset consists of:
+- **Total Images:** 5,678 (5,578 after processing)
+- **Image Dimensions:** 303×404 pixels (original), resized to 160×160 pixels for training
+- **Color Depth:** 24 bits/pixel
+- **Number of Classes:** 50 tree species
+- **Format:** Image classification directory tree structure
 
-[More Information Needed]
+### FiftyOne Dataset Structure
 
-### Out-of-Scope Use
+This dataset has been enhanced using FiftyOne with additional computed fields:
 
-<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
+**Sample Fields:**
+- `id`: Unique sample identifier (ObjectIdField)
+- `filepath`: Path to the image file (StringField)
+- `tags`: Custom tags for organizing samples (ListField)
+- `metadata`: Image metadata including dimensions and size (ImageMetadata)
+- `created_at`: Timestamp of sample creation (DateTimeField)
+- `last_modified_at`: Timestamp of last modification (DateTimeField)
+- `ground_truth`: Original tree species label (Classification)
+- `radio_embeddings`: Feature embeddings from NVIDIA RADIO v3-h model (VectorField)
+- `radio_heatmap`: Spatial attention heatmaps from RADIO model (Heatmap)
+- `siglip2_predictions`: Zero-shot predictions from SigLIP2 model (Classification)
+- `eval_simple`: Evaluation results comparing predictions to ground truth (BooleanField)
 
-[More Information Needed]
+### Data Splits
 
-## Dataset Structure
+The dataset is typically split as:
+- **Training set:** 80% of samples (approximately 4,462 images)
+- **Testing set:** 20% of samples (approximately 1,116 images)
 
-<!-- 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. -->
-
-[More Information Needed]
+K-Fold Cross Validation with k=5 has been used for robust performance evaluation.
 
 ## Dataset Creation
 
 ### Curation Rationale
 
-<!-- Motivation for the creation of this dataset. -->
+This dataset was created to address the need for automated tree species identification in forestry applications. Traditional manual identification methods are:
+- Expensive and time-consuming
+- Require expert knowledge and years of experience
+- Inconsistent in accuracy (previous studies showed expert accuracy ranging from 56.6% to 77.8%)
+- Risky, requiring experts to venture into forested areas
 
-[More Information Needed]
+The dataset focuses on bark texture because:
+1. Bark remains present throughout all seasons, unlike leaves, flowers, or fruits
+2. Bark provides characteristic identity through structural variations
+3. Bark texture is suitable for computer vision analysis
+4. Commercial and medicinal products are often derived from tree bark
 
-### Source Data
+With 50 tree species, this represents the largest bark classification dataset at the time of creation, addressing the lack of diverse, multi-class datasets in this domain.
 
-<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
+### Source Data
 
 #### Data Collection and Processing
 
-<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
-
-[More Information Needed]
+The images in BarkVN-50 were collected and curated by Truong Hoang, Vinh. The dataset contains bark texture images from 50 different tree species.
 
-#### Who are the source data producers?
+**Processing steps include:**
+- Original images: 303×404 pixels, 24 bits/pixel color depth
+- Images stored in numpy arrays after resampling and preprocessing
+- For model training, images are typically resized to 160×160 pixels
+- Data augmentation techniques are commonly applied during training
 
-<!-- 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. -->
+**Enhanced Processing (FiftyOne version):**
+- Computed embeddings using NVIDIA RADIO v3-h model for similarity search and visualization
+- Generated spatial attention heatmaps to identify discriminative bark regions
+- Applied zero-shot classification using SigLIP2 model
+- Created UMAP visualizations for embedding space exploration
 
-[More Information Needed]
+#### Who are the source data producers?
 
-### Annotations [optional]
+The original dataset was curated by **Truong Hoang, Vinh** and published on Mendeley Data. The research paper utilizing this dataset was authored by **Sahil Faizal** from the School of Computer Science and Engineering at Vellore Institute of Technology, Chennai.
 
-<!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
+### Annotations
 
 #### Annotation process
 
-<!-- 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. -->
+The dataset is organized in an image classification directory tree structure, where each subdirectory represents a tree species class. Images are labeled by their directory placement, with 50 distinct species classes.
 
-[More Information Needed]
+**Enhanced Annotations (FiftyOne version):**
+- **Ground truth labels:** Original species classifications from directory structure
+- **RADIO embeddings:** 1024-dimensional feature vectors computed using NVIDIA's RADIO v3-h model
+- **Attention heatmaps:** Spatial feature maps showing which bark regions are most discriminative
+- **SigLIP2 predictions:** Zero-shot classification predictions using the prompt "A picture of a tree bark from the species [species_name]"
+- **Evaluation metrics:** Binary correctness indicators comparing predictions to ground truth
 
 #### Who are the annotators?
 
-<!-- This section describes the people or systems who created the annotations. -->
-
-[More Information Needed]
-
-#### Personal and Sensitive Information
+The original species labels were assigned by the dataset curator, Truong Hoang, Vinh, presumably based on expert botanical knowledge. The enhanced annotations (embeddings, heatmaps, predictions) were computed algorithmically using pre-trained foundation models.
 
-<!-- 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. -->
+## Citation
 
-[More Information Needed]
+**Dataset Citation (BibTeX):**
 
-## Bias, Risks, and Limitations
-
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-
-[More Information Needed]
-
-### Recommendations
-
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-
-Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations.
-
-## Citation [optional]
-
-<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
+```bibtex
+@data{gbt4tdmttn-1,
+  author = {Truong Hoang, Vinh},
+  title = {BarkVN-50},
+  year = {2020},
+  publisher = {Mendeley Data},
+  version = {V1},
+  doi = {10.17632/gbt4tdmttn.1},
+  url = {https://data.mendeley.com/datasets/gbt4tdmttn/1}
+}
+```
 
-**BibTeX:**
+**Research Paper Citation (BibTeX):**
 
-[More Information Needed]
+```bibtex
+@article{faizal2022automated,
+  title={Automated Identification of Tree Species by Bark Texture Classification Using Convolutional Neural Networks},
+  author={Faizal, Sahil},
+  journal={arXiv preprint arXiv:2210.09290},
+  year={2022},
+  url={https://arxiv.org/abs/2210.09290v1}
+}
+```
 
 **APA:**
 
-[More Information Needed]
-
-## Glossary [optional]
-
-<!-- If relevant, include terms and calculations in this section that can help readers understand the dataset or dataset card. -->
-
-[More Information Needed]
+Truong Hoang, V. (2020). *BarkVN-50* (Version 1) [Data set]. Mendeley Data. https://doi.org/10.17632/gbt4tdmttn.1
 
-## More Information [optional]
-
-[More Information Needed]
-
-## Dataset Card Authors [optional]
-
-[More Information Needed]
+Faizal, S. (2022). Automated Identification of Tree Species by Bark Texture Classification Using Convolutional Neural Networks. *arXiv preprint arXiv:2210.09290*.
 
 ## Dataset Card Contact
 
-[More Information Needed]
+For questions about the original dataset, please refer to the [Mendeley Data repository](https://data.mendeley.com/datasets/gbt4tdmttn/1).
+
+For questions about the research paper, contact: Sahil Faizal, Vellore Institute of Technology, Chennai.