Datasets:
Tasks:
Image Classification
Modalities:
Image
Formats:
imagefolder
Languages:
English
Size:
10K - 100K
ArXiv:
Update README.md
Browse files
README.md
CHANGED
|
@@ -60,11 +60,6 @@ dataset_summary: '
|
|
| 60 |
|
| 61 |
# Dataset Card for BIOSCAN-30k
|
| 62 |
|
| 63 |
-
<!-- Provide a quick summary of the dataset. -->
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
|
| 69 |
This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 30000 samples.
|
| 70 |
|
|
@@ -93,13 +88,11 @@ session = fo.launch_app(dataset)
|
|
| 93 |
|
| 94 |
## Dataset Details
|
| 95 |
|
| 96 |
-
### Dataset Description
|
| 97 |
-
|
| 98 |
The BIOSCAN-5M dataset is a multimodal collection of over 5.15 million arthropod specimens (98% insects), curated to advance biodiversity monitoring through machine learning. It expands the earlier BIOSCAN-1M dataset by including high-resolution images, DNA barcode sequences, taxonomic labels (phylum to species), geographical locations, specimen size data, and Barcode Index Numbers (BINs). Designed for both closed-world (known species) and open-world (novel species) scenarios, it supports tasks like taxonomic classification, clustering, and multimodal learning.
|
| 99 |
|
| 100 |
**This dataset is a randomly chosen subset of 30,000 samples across all splits from the Cropped 256 dataset**
|
| 101 |
|
| 102 |
-
Key Features:
|
| 103 |
|
| 104 |
* Images: 5.15M high-resolution microscope images (1024×768px) with cropped/resized versions.
|
| 105 |
|
|
@@ -111,6 +104,7 @@ Key Features:
|
|
| 111 |
|
| 112 |
* Size Metadata: Pixel count, area fraction, and scale factor for specimens.
|
| 113 |
|
|
|
|
| 114 |
|
| 115 |
- **Curated by:** International consortium led by the Centre for Biodiversity Genomics, University of Guelph, University of Waterloo, and Simon Fraser University.
|
| 116 |
- **Funded by :** Government of Canada’s New Frontiers in Research Fund (NFRF), CFI Major Science Initiatives, Walder Foundation, NVIDIA Academic Grant.
|
|
|
|
| 60 |
|
| 61 |
# Dataset Card for BIOSCAN-30k
|
| 62 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
|
| 64 |
This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 30000 samples.
|
| 65 |
|
|
|
|
| 88 |
|
| 89 |
## Dataset Details
|
| 90 |
|
|
|
|
|
|
|
| 91 |
The BIOSCAN-5M dataset is a multimodal collection of over 5.15 million arthropod specimens (98% insects), curated to advance biodiversity monitoring through machine learning. It expands the earlier BIOSCAN-1M dataset by including high-resolution images, DNA barcode sequences, taxonomic labels (phylum to species), geographical locations, specimen size data, and Barcode Index Numbers (BINs). Designed for both closed-world (known species) and open-world (novel species) scenarios, it supports tasks like taxonomic classification, clustering, and multimodal learning.
|
| 92 |
|
| 93 |
**This dataset is a randomly chosen subset of 30,000 samples across all splits from the Cropped 256 dataset**
|
| 94 |
|
| 95 |
+
#### Key Features:
|
| 96 |
|
| 97 |
* Images: 5.15M high-resolution microscope images (1024×768px) with cropped/resized versions.
|
| 98 |
|
|
|
|
| 104 |
|
| 105 |
* Size Metadata: Pixel count, area fraction, and scale factor for specimens.
|
| 106 |
|
| 107 |
+
### Dataset Description
|
| 108 |
|
| 109 |
- **Curated by:** International consortium led by the Centre for Biodiversity Genomics, University of Guelph, University of Waterloo, and Simon Fraser University.
|
| 110 |
- **Funded by :** Government of Canada’s New Frontiers in Research Fund (NFRF), CFI Major Science Initiatives, Walder Foundation, NVIDIA Academic Grant.
|