fa0311/oita-ken-strawberries-mobilenet

oita-ken-strawberries-mobilenet

This model is designed for grading the ベリーツ (Beriitsu), a premium strawberry variety cultivated in Oita Prefecture, Japan.

It is a fine-tuned version of MobileNet SSD v2, optimized for strawberry quality classification.

The model has been trained on an augmented dataset derived from oita-ken-strawberries, using image_data_augmentation.

It performs 4-class classification based on an 40,000-image dataset.

Learning

https://github.com/fa0311/pytorch-ssd-forked

python3 train_ssd.py \
    --pretrained-ssd=models/mb2-ssd-lite-mp-0_686.pth \
    --net=mb2-ssd-lite \
    --dataset-type=voc \
    --data=data/voc \
    --model-dir=models/output \
    --num-workers=8 \
    --batch-size=16 \
    --num-epochs=56 \
    --t-0=8 \
    --validation-epochs=1 \
    --lr=0.0001 \
    --extra-layers-lr=0.001 \
    --base-net-lr=0.000001 \
    --weight-decay=0.000001 \
    --validation-mean-ap \
    --scheduler=cosine-warmrestart \
    --no-augment

Related Resources

• Dataset
• Model
• Inference Space