med-herb-lens/model-3-4661eff4

model 3

An image classifier trained by a Med Herb Lens user on their own device, and shared under CC BY 4.0.

What it identifies (10 species)

• Cyperus iria
• Echinochloa colonum
• Imperata cylindrica
• Ipomoea aquatica
• Mimosa pigra
• Nelumbo nucifera
• Phragmites karka
• Phyllanthus reticulatus
• Praxelis clematidea
• Tropaeolum majus

How to use it

• In Med Herb Lens: download model.tflite, then Train → Import a model. Made in Med Herb Lens, it can go on learning there: add species or photos and train again.
• Anywhere TensorFlow Lite runs: a 224 × 224 RGB image with values from 0 to 1 in; one probability per species out. The species list is inside the file (labels.txt in its metadata).

How it was made

A final layer trained in Med Herb Lens, on the device, on top of the dinov2_vits14 MediaPipe image embedder (Apache 2.0).

Caution

Like any model, it can be wrong, and its species names were typed by its sharer. Never eat a plant or use it as medicine because of what a model says.

Credit and citation

Shared by a Med Herb Lens user (shared model 4661eff4-e00a-484d-accb-a1b39e4e5b74). If you use it, please cite:

Tran, T. P., Ud Din, F., Brankovic, L., Sanin, C., & Hester, S. M. (2025). Med Herb Lens: A prototype AI app for medicinal plant identification. Procedia Computer Science, 270, 2603–2612. https://doi.org/10.1016/j.procs.2025.09.382

Tran, T. P., Ud Din, F., Brankovic, L., Sanin, C., & Hester, S. M. (2026). Resource-efficient continual learning for medicinal plant identification: A periodic retraining approach for edge-deployed agricultural IoT applications. IoT, 7(3), 57. https://doi.org/10.3390/iot7030057