YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.


language:
- en
tags:
- image-classification
- bangla-food
- bangladeshi-food
- food-recognition
- efficientnetv2
library_name: transformers
base_model: timm/tf_efficientnetv2_s.in21k_ft_in1k
metrics:
- accuracy
- f1
---

# Bangla Food Recognizer - EfficientNetV2-S

Fine-tuned from EfficientNetV2-S `timm/tf_efficientnetv2_s.in21k_ft_in1k` for Bangladeshi food image recognition.

## Training Sources

- `tanziro/bangladeshi-food-1041`
- DeshiFoodBD, Mendeley Data DOI `10.17632/tczzndbprx.1`
- FoodBD polygon-annotated meal images, Mendeley Data DOI `10.17632/xh3ghf3jbg.2`
- BD-Food33, Mendeley Data DOI `10.17632/8wt48bjmx3.1`

## Evaluation

```json
{

"validation": { "validation_loss": 2.2158524990081787, "validation_accuracy": 0.6118909991742362, "validation_macro_f1": 0.2587394309318114, "validation_weighted_f1": 0.5640698507031976, "validation_top3_accuracy": 0.7448389760528489 }, "test": { "test_loss": 2.483065605163574, "test_accuracy": 0.5770750988142292, "test_macro_f1": 0.2316382332791406, "test_weighted_f1": 0.5353572882969004, "test_top3_accuracy": 0.7114624505928854 } } ```

## Intended Use

Exploratory Bangladeshi food recognition from images. This is not a nutrition, medical, allergy, or food-safety system. Check original dataset licences and terms before commercial use.
Downloads last month
3
Safetensors
Model size
21.4M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support