edoggy commited on
Commit
c6ea101
·
verified ·
1 Parent(s): b110076

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. README.md +39 -0
  2. crop_classifier_v1.joblib +3 -0
  3. metrics.json +80 -0
README.md ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-sa-4.0
3
+ library_name: sklearn
4
+ tags: [agriculture, crop-type, crop-mapping, remote-sensing, alphaearth, tabular-classification]
5
+ datasets: [edoggy/sugarcane-satellite-embeddings]
6
+ ---
7
+
8
+ # Crop-type classifier on Satellite Embeddings (v1)
9
+
10
+ One gradient-boosted model (sklearn HistGradientBoosting) that maps a 64-dim Google Satellite Embedding
11
+ vector of a cropland pixel to a crop type. Trained on USDA CDL labels (Louisiana, Florida, Texas, 2018–2024)
12
+ and MapBiomas Collection 10 labels (São Paulo, 2020, 2023). Classes: sugarcane, soybeans, corn, cotton, rice,
13
+ winter_wheat, peanuts, sorghum, citrus, hay, sod, other_tree, pasture, forest_plantation, other.
14
+
15
+ ## Results — US 2024, trained on 2018–2023 only
16
+ Accuracy 0.823, weighted F1 0.858. Per-class F1: sugarcane 0.94, corn 0.90, soybeans 0.89, cotton 0.86,
17
+ rice 0.85, winter wheat 0.83, peanuts 0.80, citrus 0.76, sorghum 0.68, hay 0.49, sod 0.17.
18
+
19
+ ## Known confusions (honest limits)
20
+ - US: sugarcane→sorghum (both tall C4 grasses), citrus→sod and citrus→hay (Florida sod farms between groves),
21
+ hay↔sod↔other (grassy classes CDL itself labels inconsistently).
22
+ - Brazil (SP 2023 holdout): sugarcane→pasture, ~10% of cane pixels. Cane on marginal land and degraded
23
+ pasture overlap in the embedding.
24
+
25
+ ## Scope
26
+ - Class proportions reflect the sampling: sugarcane is oversampled (50% of rows); other crops appear at their
27
+ natural frequency among cropland. Rare classes (<1,500 rows) are folded into `other`.
28
+ - Embeddings are strongly region/season-specific; this model is calibrated to the trained regions. Elsewhere,
29
+ label a few thousand local pixels and retrain (see the sugarcane classifier card for the dose curve).
30
+ - Apply only to pixels already known to be cropland.
31
+ - Not a price or supply predictor: annual embeddings arrive after official acreage reports.
32
+
33
+ ## Usage
34
+ ```python
35
+ import joblib; from huggingface_hub import hf_hub_download
36
+ m = joblib.load(hf_hub_download("edoggy/crop-classifier-v1", "crop_classifier_v1.joblib"))
37
+ m.predict(embeddings_64d) # returns class names
38
+ ```
39
+ Embeddings: Google Satellite Embedding V1 (CC-BY-4.0). Labels: USDA CDL (public domain), MapBiomas C10 (CC-BY-4.0).
crop_classifier_v1.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16ea236584060f3dbc72a150076bd256727895cc5084ee33160bb65e628e5fe5
3
+ size 682112
metrics.json ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "classes": [
3
+ "citrus",
4
+ "corn",
5
+ "cotton",
6
+ "forest_plantation",
7
+ "hay",
8
+ "other",
9
+ "other_tree",
10
+ "pasture",
11
+ "peanuts",
12
+ "rice",
13
+ "sod",
14
+ "sorghum",
15
+ "soybeans",
16
+ "sugarcane",
17
+ "winter_wheat"
18
+ ],
19
+ "cdl_map": {
20
+ "45": "sugarcane",
21
+ "5": "soybeans",
22
+ "1": "corn",
23
+ "3": "rice",
24
+ "24": "winter_wheat",
25
+ "2": "cotton",
26
+ "37": "hay",
27
+ "212": "citrus",
28
+ "72": "citrus",
29
+ "61": "fallow",
30
+ "59": "sod",
31
+ "26": "dbl_wheat_soy",
32
+ "4": "sorghum",
33
+ "10": "peanuts",
34
+ "215": "avocado",
35
+ "71": "other_tree",
36
+ "47": "vegetables",
37
+ "216": "vegetables",
38
+ "221": "strawberries"
39
+ },
40
+ "mapbiomas_map": {
41
+ "20": "sugarcane",
42
+ "39": "soybeans",
43
+ "41": "other_temp_crop",
44
+ "40": "rice",
45
+ "62": "cotton",
46
+ "46": "coffee",
47
+ "47": "citrus",
48
+ "48": "other_perennial",
49
+ "9": "forest_plantation",
50
+ "15": "pasture"
51
+ },
52
+ "us_2024_holdout": {
53
+ "accuracy": 0.823,
54
+ "weighted_f1": 0.858,
55
+ "macro_f1": 0.695,
56
+ "per_class_f1": {
57
+ "sugarcane": 0.94,
58
+ "corn": 0.9,
59
+ "soybeans": 0.89,
60
+ "cotton": 0.86,
61
+ "rice": 0.85,
62
+ "winter_wheat": 0.83,
63
+ "peanuts": 0.8,
64
+ "citrus": 0.76,
65
+ "sorghum": 0.68,
66
+ "hay": 0.49,
67
+ "sod": 0.17
68
+ }
69
+ },
70
+ "top_confusions_us": [
71
+ "sugarcane->sorghum",
72
+ "citrus->sod",
73
+ "citrus->hay",
74
+ "hay->sod"
75
+ ],
76
+ "top_confusions_brazil": [
77
+ "sugarcane->pasture (~10% of cane)",
78
+ "sugarcane->soybeans"
79
+ ]
80
+ }