ixrbhii/coconut-disease-convnext

Coconut leaf disease classifier

Tells which of 11 classes a coconut leaf photo shows: Bud Root Dropping, Bud Rot, Coconut Caterpillar Infestation (caterpillars), Coconut Caterpillar Infestation (leaflet damage), Gray Leaf Spot, Healthy, Leaf Rot, Stem Bleeding, Weligama Coconut Leaf Wilt — drying of leaflets, Weligama Coconut Leaf Wilt — flaccidity, Weligama Coconut Leaf Wilt — yellowing.

ConvNeXt-Small (convnext_small, from convnext_small.fb_in22k_ft_in1k), 224 × 224 RGB, 99 MB (half precision). 99.9 % test accuracy, macro-F1 0.998, on 1,402 held-out photos. Test photos come from the same datasets the model learned from, so real field photos score lower.

Class	Recall	F1	Test photos
Bud Root Dropping	96.9%	0.984	64
Bud Rot	100.0%	1.000	59
Coconut Caterpillar Infestation (caterpillars)	100.0%	1.000	124
Coconut Caterpillar Infestation (leaflet damage)	100.0%	1.000	99
Gray Leaf Spot	100.0%	1.000	267
Healthy	100.0%	1.000	50
Leaf Rot	100.0%	0.995	209
Stem Bleeding	100.0%	1.000	126
Weligama Coconut Leaf Wilt — drying of leaflets	100.0%	1.000	135
Weligama Coconut Leaf Wilt — flaccidity	100.0%	1.000	134
Weligama Coconut Leaf Wilt — yellowing	100.0%	1.000	135

Use

import timm, torch
from PIL import Image

model = timm.create_model("hf-hub:ixrbhii/coconut-disease-convnext", pretrained=True).eval()
cfg = timm.data.resolve_data_config({}, model=model)
x = timm.data.create_transform(**cfg)(Image.open("leaf.jpg").convert("RGB")).unsqueeze(0)
p = model(x).softmax(-1)[0]
labels = model.pretrained_cfg["label_names"]
for i in p.argsort(descending=True)[:3]:
    print(labels[i], f"{100 * p[i]:.1f} %")

The training images were resized to 224 × 224 without cropping; timm's transform above (resize + 224 centre crop) gives the same result for square photos. label_descriptions in config.json maps each name to the dataset's original label. Best results: one leaf filling most of the photo, in daylight and in focus. Below about 45 % confidence, ask for a better photo.

In an app: LULC Fetch (Agri ▸ Diagnose crop disease) recognises the crop first and then runs the right disease model, with a photo-quality check and a disease map from geotagged photos.

Related

• All 44 models (2 crop detectors, 42 crops) as separate repositories: the collection
• All 44 models in one repository: ixrbhii/multicrop-disease-models
• Symptoms, treatment and pests for each disease: ixrbhii/crop-disease-qa

Training data and licence

Trained on public leaf-disease photo datasets, including PlantVillage (CC0) and MangoLeafBD (CC-BY-4.0). Released under CC-BY-4.0: please credit "Multi-crop disease models, IXR (agritechixr)" and the training datasets. A diagnosis supports, but doesn't replace, a local agriculture expert.