Rank 13/835 Top(1.56%)
The Aarhus University Signal Processing group, in collaboration with University of Southern Denmark, has recently released a dataset containing images of approximately 960 unique plants belonging to 12 species at several growth stages.
The aim of this project is to build a classifier that can identify plant species from an image.
List of 12 different species:
- Black-grass
- Charlock
- Cleavers
- Common Chickweed
- Common wheat
- Fat Hen
- Loose Silky-bent
- Maize
- Scentless Mayweed
- Shepherds Purse
- Small-flowered Cranesbill
- Sugar beet
Submissions are evaluated on Mean Score, which at Kaggle is a micro-average F1-score.
-
Segmentation
-
Balancing Dataset
- K-Means Clustering
- K-Neirest Neighbours Classifier
- Convolutional Neural Networks
- EfficientNetB3
- Xception
- Resnet-50
- Inception-ResNet-v2
- Vision Transformer
- Weighted Average Ensemble
- Binary Classifier Error Correction
- Test-Time-Augmentation
Our final solution is the weighted average ensemble that combines the probabilities of the Xception, EfficientNetB3 and InceptionResNet-v2 models using a weighted average aggregator
The final probabilities are given by
P(X) = 0.1 * EfficientNetB3(X) + 0.5 * Inception-ResNet-v2(X) + 0.4 * Xception(X)
