andyvhuynh
commited on
Commit
•
e9c4bbc
1
Parent(s):
000326f
Update README.md
Browse files
README.md
CHANGED
@@ -59,4 +59,48 @@ dataset_info:
|
|
59 |
sequence:
|
60 |
sequence:
|
61 |
sequence: int64
|
62 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
59 |
sequence:
|
60 |
sequence:
|
61 |
sequence: int64
|
62 |
+
---
|
63 |
+
|
64 |
+
Nature Multi-View (NMV) Dataset Datacard
|
65 |
+
Dataset Overview
|
66 |
+
|
67 |
+
Name: Nature Multi-View Dataset (NMV)
|
68 |
+
Total Observations: 3,921,026
|
69 |
+
Total Images: Over 3 million ground-level and aerial image pairs
|
70 |
+
Species Covered: Over 6,000 native and introduced plant species
|
71 |
+
Geographic Focus: California, USA
|
72 |
+
Data Description
|
73 |
+
|
74 |
+
Ground-Level Images:
|
75 |
+
Sourced from iNaturalist open data on AWS.
|
76 |
+
Filters applied:
|
77 |
+
Vascular plants
|
78 |
+
Within California state boundaries
|
79 |
+
Observations dated from January 1, 2011, to September 27, 2023
|
80 |
+
Geographic uncertainty < 120 meters
|
81 |
+
Research-grade or in need of ID (excluding casual observations)
|
82 |
+
Availability of corresponding remote sensing imagery
|
83 |
+
Overlap with bio-climatic variables
|
84 |
+
Aerial Images:
|
85 |
+
Sourced from the 2018 National Agriculture Imagery Program (NAIP).
|
86 |
+
RGB-Infrared images, 256x256 pixels, 60 cm-per-pixel resolution.
|
87 |
+
Centered on the latitude and longitude of the iNaturalist observation.
|
88 |
+
Dataset Splits
|
89 |
+
|
90 |
+
Training Set:
|
91 |
+
Full Training Set: 1,755,602 observations, 3,307,025 images
|
92 |
+
Labeled Training Sets:
|
93 |
+
20%: 334,383 observations, 390,908 images
|
94 |
+
5%: 93,708 observations, 97,727 images
|
95 |
+
1%: 19,371 observations, 19,545 images
|
96 |
+
0.25%: 4,878 observations, 4,886 images
|
97 |
+
Validation Set:
|
98 |
+
150,555 observations, 279,114 images
|
99 |
+
Test Set:
|
100 |
+
182,618 observations, 334,887 images
|
101 |
+
Characteristics and Challenges
|
102 |
+
|
103 |
+
Long-Tail Distribution: The dataset exhibits a long-tail distribution common in natural world settings, making it a realistic benchmark for machine learning applications.
|
104 |
+
Geographic Bias: Reflects the geographic bias of citizen science data, with more observations from densely populated and visited regions like urban areas and National Parks.
|
105 |
+
Many-to-One Pairing: Multiple ground-level images are paired to the same aerial image, which may require special handling in machine learning models.
|
106 |
+
Self-Supervised Learning: Designed to encourage the development of self-supervised machine learning methods, unrestricted by label balance.
|