Plant Leaf Species Classifier — 4-Model Comparison
Four ImageNet-pretrained backbones fine-tuned on a real-world (field-condition) plant leaf dataset, compared on accuracy vs. computational cost.
| Model | Params | Test Accuracy (%) | Avg. Inference (ms/img) |
|---|---|---|---|
| MobileNetV4-Conv-Small | 2,503,272 | 99.56 | 1.026 |
| MobileNetV4-Conv-Medium | 8,444,760 | 98.67 | 1.254 |
| EfficientNetV2-S | 20,187,736 | 100.0 | 2.399 |
| ConvNeXt-Tiny | 27,826,280 | 100.0 | 33.36 |
Classes: Apple, Berry, Fig, Guava, Orange, Palm, Persimmon, Tomato
Each <key>_leaf_classifier.pth is a state_dict() for the corresponding
timm model (see model_registry.json for the exact timm model name to
reconstruct the architecture with timm.create_model(timm_name, pretrained=False, num_classes=8)).
- Downloads last month
- -