Mapúa Institute of Technology • Model Demo

🌴 Lightweight Hybrid ViT-CNN for Enhanced Coconut Disease Detection Using MobileViTv3

Authors: Jaden Louie E. Laudes, Kristian A. Soner, Psalmwel Kyle M. Conciso

Model Selection & Specifications

Select one of the trained hybrid architectures to load into the client-side execution session.

Model Architecture MobileViTv3 (Hybrid)
Parameter Count 2.30 M
FLOPs (Inference Complexity) 0.18 GFLOPs
Validation F1-Score (Macro) 98.70%
ONNX File Size 8.40 MB

Image Input Source

Select or drag & drop leaf image for inference

Supported formats: PNG, JPG, JPEG (auto-cropped to square input resolution)
Camera stream inactive

Standardized evaluation samples validated by botanical experts:

Bud Rot Sample
Bud Rot
Gray Leaf Spot Sample
Gray Leaf Spot
Leaf Rot Sample
Leaf Rot
Stem Bleeding Sample
Stem Bleeding

Visual Inspection Area

No image loaded. Upload an image, use the camera, or select an evaluation sample to run model evaluation.

Quantitative Predictions

Ready. Please provide an input image.