hylin16/food-recognition-food11-effnet-b4-cbam-300

efficientnet_b4_cbam — Food-11

Food image classifier trained with food_recognition. efficientnet_b4_cbam is a torchvision backbone with a CBAM block (channel then spatial attention) between the features and the classification head.

Measured performance

Metric	Value
Validation accuracy	94.85%
Macro F1	0.9481
Classes	11
Input resolution	300px
Checkpoint epoch	30
Training wall clock	49 min
Hardest / easiest class	02 (F1 0.855) / 06 (F1 1.000)

These figures come from this run's own metrics.json, computed over the full validation split. The repository re-scores every published checkpoint through a separate evaluation path and requires the two to agree to six decimal places.

Usage

pip install git+https://github.com/linhongyu510/food_recognition.git

from huggingface_hub import hf_hub_download
from food_recognition import load_predictor

checkpoint = hf_hub_download("hylin16/food-recognition-food11-effnet-b4-cbam-300", "best.pt")
predictor = load_predictor(checkpoint)

prediction = predictor.predict("photo.jpg", topk=3)
print(prediction.label, prediction.topk)

The checkpoint embeds its architecture, input resolution and class names, so nothing needs to be specified again at load time.

To see which pixels drove a prediction:

food-recognition-gradcam --checkpoint best.pt --input photo.jpg --side-by-side

Training

• Architecture: efficientnet_b4_cbam
• Input: 300px
• Epochs: 30
• Batch size: 32
• Learning rate: 0.0005
• Scheduler: cosine
• Fully supervised (no pseudo-labelling)
• Source commit: baf1b5e

Limitations

• Trained on Food-11 only. Accuracy on dishes, cuisines or photographic conditions outside that distribution is unmeasured, and the model will still return a confident label for an image containing no food at all.
• Predictions are not a nutrition, allergen or food-safety judgement. Do not use them where a misclassification carries a health risk.
• The reported figure is validation accuracy on a public benchmark split, which is an optimistic estimate of field performance.