Rebrand to HUGO-Bench: Hierarchical Unsupervised Grouping of Organisms Benchmark

README.md

@@ -1,4 +1,4 @@
----
+---
 license: mit
 task_categories:
   - image-classification
@@ -9,20 +9,34 @@ tags:
   - camera-trap
   - animals
   - vision
+  - zero-shot
+  - clustering
+  - benchmark
+  - HUGO-Bench
 size_categories:
   - 100K<n<1M
 ---
-# ViT Animal Dataset
 
-This dataset contains 139,111 cropped images of 60 animal species organized Aves and Mammalia.
-The raw image data was derived from 23 different projects across LILA BC, filtered and cropped to only include the detected species. 
-The data has been used to benchmark different ViT models, combined with dimension reduction & clustering methods. 
+# HUGO-Bench 
+
+**Hierarchical Unsupervised Grouping of Organisms Benchmark**
+
+A comprehensive benchmark dataset for evaluating zero-shot clustering of wildlife camera trap images using Vision Transformer embeddings.
+
+## Overview
+
+HUGO-Bench contains **139,111 expert-validated cropped images** of 60 animal species (30 birds, 30 mammals), derived from 23 camera trap projects across [LILA BC](https://lila.science/). The dataset enables benchmarking of Vision Transformer models for unsupervised species-level clustering without requiring labeled training data.
 
 ## Dataset Details
-- **Format**: Parquet (optimized for streaming)
-- **Total Images**: 139,111
-- **Species Count**: 60 species (30 birds, 30 mammals)
-- **License**: MIT
+
+| Property | Value |
+|----------|-------|
+| **Format** | Parquet (optimized for streaming) |
+| **Total Images** | 139,111 |
+| **Species Count** | 60 species (30 birds, 30 mammals) |
+| **Validated** | 138,024 (99.2%) |
+| **Uncertain** | 1,087 (0.8%) |
+| **License** | MIT |
 
 ### Splits
 - `aves`: 73,528 images of 30 bird species
@@ -45,7 +59,7 @@ Filenames follow the pattern: `{species-name}_{number}.jpg` or `uncertain_{speci
 from datasets import load_dataset
 
 # Load with streaming enabled
-dataset = load_dataset("AI-EcoNet/ViT_animal_data", split="aves", streaming=True)
+dataset = load_dataset("AI-EcoNet/HUGO-Bench", split="aves", streaming=True)
 
 # Iterate through samples
 for sample in dataset:
@@ -60,7 +74,7 @@ for sample in dataset:
 from datasets import load_dataset
 
 # Stream and filter for specific species
-ds = load_dataset("AI-EcoNet/ViT_animal_data", split="aves", streaming=True)
+ds = load_dataset("AI-EcoNet/HUGO-Bench", split="aves", streaming=True)
 crows_only = ds.filter(lambda x: x['filename'].startswith('american-crow'))
 
 for sample in crows_only:
@@ -72,46 +86,10 @@ for sample in crows_only:
 from datasets import load_dataset
 
 # Load only first 1000 images from aves
-dataset = load_dataset("AI-EcoNet/ViT_animal_data", split="aves[:1000]")
+dataset = load_dataset("AI-EcoNet/HUGO-Bench", split="aves[:1000]")
 
 # Load 10% of mammals
-dataset = load_dataset("AI-EcoNet/ViT_animal_data", split="mammals[:10%]")
-```
-
-### Load exactly 200 images per species (validated only)
-```python
-from datasets import load_dataset
-from collections import defaultdict
-
-# Load with streaming
-ds = load_dataset("AI-EcoNet/ViT_animal_data", split="aves", streaming=True)
-
-# Collect 200 validated images per species
-species_list = ['american-crow', 'australian-magpie', 'kea']
-species_counts = defaultdict(int)
-species_samples = defaultdict(list)
-
-for sample in ds:
-    filename = sample['filename']
-    
-    # Skip uncertain images
-    if filename.startswith('uncertain_'):
-        continue
-    
-    # Extract species name
-    species = filename.rsplit('_', 1)[0]
-    
-    # Collect if we need more
-    if species in species_list and species_counts[species] < 200:
-        species_samples[species].append(sample)
-        species_counts[species] += 1
-        
-        # Stop when we have 200 of each
-        if all(species_counts[s] >= 200 for s in species_list):
-            break
-
-print(f"Collected: {dict(species_counts)}")
-# Output: {'american-crow': 200, 'australian-magpie': 200, 'kea': 200}
+dataset = load_dataset("AI-EcoNet/HUGO-Bench", split="mammals[:10%]")
 ```
 
 ### Full download
@@ -119,16 +97,16 @@ print(f"Collected: {dict(species_counts)}")
 from datasets import load_dataset
 
 # Download full dataset
-dataset = load_dataset("AI-EcoNet/ViT_animal_data")
+dataset = load_dataset("AI-EcoNet/HUGO-Bench")
 
 # Access by split
 aves_data = dataset['aves']
 mammals_data = dataset['mammals']
 ```
 
-### Complete Species List
+## Complete Species List
 
-#### Aves (Birds) - 30 Species, 73,528 Images
+### Aves (Birds) - 30 Species, 73,528 Images
 
 | Species | Total | Validated | Uncertain |
 |---------|-------|-----------|-----------|
@@ -139,7 +117,7 @@ mammals_data = dataset['mammals']
 | Blue whistling thrush | 2,157 | 2,128 | 29 |
 | Brown quail | 2,662 | 2,652 | 10 |
 | Chicken | 1,647 | 1,629 | 18 |
-| Common chafflinch | 2,384 | 2,370 | 14 |
+| Common chaffinch | 2,384 | 2,370 | 14 |
 | Common myna | 2,707 | 2,704 | 3 |
 | Dunnock | 2,066 | 2,055 | 11 |
 | European starling | 1,876 | 1,867 | 9 |
@@ -164,7 +142,7 @@ mammals_data = dataset['mammals']
 | Wild turkey | 2,057 | 2,057 | 0 |
 | Yellow-eyed penguin | 2,817 | 2,815 | 2 |
 
-#### Mammals - 30 Species, 65,583 Images
+### Mammals - 30 Species, 65,583 Images
 
 | Species | Total | Validated | Uncertain |
 |---------|-------|-----------|-----------|
@@ -199,19 +177,41 @@ mammals_data = dataset['mammals']
 | White-nosed coati | 1,996 | 1,982 | 14 |
 | Wolf | 2,307 | 2,307 | 0 |
 
-**Note**: "Uncertain" indicates cases where manual validation was uncertain of the intitial label.
+**Note**: "Uncertain" indicates cases where manual validation was uncertain of the initial label.
+
+## Benchmark Results
+
+Using HUGO-Bench, we evaluated combinations of 5 ViT models  5 dimensionality reduction methods  4 clustering algorithms. Key findings:
+
+| Configuration | V-Measure | Notes |
+|---------------|-----------|-------|
+| DINOv3 + t-SNE + Hierarchical (K=30) | **0.958** | Best supervised |
+| DINOv3 + t-SNE + HDBSCAN | **0.943** | Best unsupervised |
+| DINOv3 + t-SNE + HDBSCAN | 33-37 clusters | ~10% overestimate of true species count |
+
+For complete benchmarking results, see our paper and the [interactive demo](https://hugomarkoff.github.io/animal_visual_transformer/).
 
 ## Citation
 
-If you use this dataset, please cite appropriately. Each species folder contains the original image source and datasets from LILA BC, which can be cited seperately. 
+If you use HUGO-Bench, please cite:
 
-```bibtex
-@dataset{markoff_vit_animal_data_2026,
+`ibtex
+@dataset{hugo_bench_2026,
   author       = {Markoff, Hugo and Bengtson, Stefan Hein and Ørsted, Michael},
-  title        = {ViT animal data: Expert-validated camera-trap cropped detections for zero-shot clustering},
+  title        = {HUGO-Bench: Hierarchical Unsupervised Grouping of Organisms Benchmark},
   year         = {2026},
   publisher    = {Hugging Face Datasets},
-  url          = {https://huggingface.co/datasets/AI-EcoNet/ViT_animal_data},
-  note         = {Each species folder lists original image sources from LILA BC; please cite those datasets as indicated. "Uncertain" denotes expert-ambiguous labels.}
+  url          = {https://huggingface.co/datasets/AI-EcoNet/HUGO-Bench},
+  note         = {Expert-validated camera-trap images for zero-shot clustering. Image sources from LILA BC.}
 }
-```
+`
+
+## Links
+
+-  [Interactive Demo](https://hugomarkoff.github.io/animal_visual_transformer/)
+-  [Benchmarking Code](https://github.com/hugomarkoff/animal_visual_transformer)
+-  Paper: *Vision Transformers for Zero-Shot Clustering of Animal Images: A Comparative Benchmarking Study*
+
+## Acknowledgments
+
+Raw image data derived from 23 projects on [LILA BC](https://lila.science/) (Labeled Information Library of Alexandria: Biology and Conservation). Please cite the original datasets as appropriate.