Swin-finetuned Food101 Open-source Food Image Classification Model - Accurately Identify Food Categories with a 92.1% Accuracy Rate

Swin Finetuned Food101

Developed by aspis

A food image classification model fine-tuned based on the Swin Transformer architecture, achieving 92.1% accuracy on the Food101 dataset

Image Classification

Transformers

Open Source License:Apache-2.0 #Food Image Classification #High-precision Recognition #Swin Transformer

Downloads 19

Release Time : 6/9/2022

Model Overview

This model is a fine-tuned version of microsoft/swin-base-patch4-window7-224 on the food101 dataset, specifically designed for food image classification tasks.

Model Features

High Accuracy

Achieves 92.1% accuracy on the Food101 validation set

Based on Swin Transformer

Utilizes the advanced Swin Transformer architecture with excellent visual feature extraction capabilities

Fine-tuned Optimization

Fine-tuned specifically on the Food101 dataset to optimize food classification performance

Model Capabilities

Food Image Classification

Multi-category Image Recognition

Use Cases

Food Recognition

Dish Recognition in Catering Industry

Used in restaurants for automatic dish recognition to assist ordering systems

Can accurately identify 101 common food categories

Healthy Diet Analysis

Helps users identify food types to assist in nutritional analysis

🚀 swin-finetuned-food101

This model is a fine - tuned version of [microsoft/swin - base - patch4 - window7 - 224](https://huggingface.co/microsoft/swin - base - patch4 - window7 - 224) on the food101 dataset. It can effectively classify images in the food101 dataset, providing high - accuracy image classification results.

🚀 Quick Start

This model is ready to use for image classification tasks. You can load it using the Hugging Face Transformers library and start making predictions.

📚 Documentation

Model Information

Property	Details
Model Type	Fine - tuned version of microsoft/swin - base - patch4 - window7 - 224
Training Data	food101
Metrics	accuracy

Model Performance

The model achieves the following results on the evaluation set:

Loss: 0.2772
Accuracy: 0.9210

The detailed performance on different tasks and datasets is as follows:

Task: Image Classification on food101 (default args)
- Accuracy: 0.9210297029702971
Task: Image Classification on food101 (validation split, default config)
- Accuracy: 0.9135841584158416 (verified)
- Precision Macro: 0.9151645786633058 (verified)
- Precision Micro: 0.9135841584158416 (verified)
- Precision Weighted: 0.915164578663306 (verified)
- Recall Macro: 0.9135841584158414 (verified)
- Recall Micro: 0.9135841584158416 (verified)
- Recall Weighted: 0.9135841584158416 (verified)
- F1 Macro: 0.9138785016966742 (verified)
- F1 Micro: 0.9135841584158415 (verified)
- F1 Weighted: 0.9138785016966743 (verified)
- Loss: 0.30761435627937317 (verified)

Training Procedure

Training Hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e - 05
train_batch_size: 16
eval_batch_size: 16
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 64
optimizer: Adam with betas=(0.9, 0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 3

Training Results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.5077	1.0	1183	0.3851	0.8893
0.3523	2.0	2366	0.3124	0.9088
0.1158	3.0	3549	0.2772	0.9210

Framework Versions

Transformers 4.19.2
Pytorch 1.11.0+cu113
Datasets 2.2.2
Tokenizers 0.12.1

📄 License

This model is released under the Apache 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご