DeiT-tiny Image Classification Model - An open-source lightweight tool with a fine-tuning validation accuracy of 98.56%

Deit Tiny Patch16 224 Finetuned Main Gpu 20e Final

Developed by Gokulapriyan

Lightweight image classification model based on DeiT-tiny architecture, achieving 98.56% validation accuracy after fine-tuning on a custom image dataset

Image Classification

Transformers

Open Source License:Apache-2.0 #High-precision image classification #Lightweight vision model #Transfer learning optimization

Downloads 15

Release Time : 4/12/2023

Model Overview

This model is a fine-tuned version of facebook/deit-tiny-patch16-224, specifically designed for image classification tasks and performs excellently on evaluation sets.

Model Features

High accuracy

Achieves 98.56% classification accuracy on validation set

Lightweight architecture

Based on DeiT-tiny's lightweight design, suitable for resource-constrained environments

Efficient fine-tuning

Requires only 20 training epochs to achieve high performance

Model Capabilities

Image classification

Visual feature extraction

Use Cases

Computer vision

Object recognition

Classify and identify objects in images

Achieves 98.56% accuracy on test data

Quality inspection

Automatic quality detection for products on industrial production lines

🚀 deit-tiny-patch16-224-finetuned-main-gpu-20e-final

This model is a fine - tuned version of facebook/deit-tiny-patch16-224 on the imagefolder dataset. It offers high - accuracy image classification capabilities, achieving excellent results on the evaluation set.

🚀 Quick Start

This model is a fine - tuned version of facebook/deit-tiny-patch16-224 on the imagefolder dataset. It achieves the following results on the evaluation set:

Loss: 0.0420
Accuracy: 0.9856

✨ Features

High Accuracy: Achieves an accuracy of 0.9856 on the evaluation set.
Fine - tuned: Based on the pre - trained facebook/deit-tiny-patch16-224 model, fine - tuned on the imagefolder dataset.

📚 Documentation

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e - 05
train_batch_size: 32
eval_batch_size: 32
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 128
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.6047	1.0	551	0.6283	0.7111
0.431	2.0	1102	0.3962	0.8366
0.352	3.0	1653	0.2620	0.8953
0.2682	4.0	2204	0.1814	0.9318
0.2533	5.0	2755	0.1564	0.9396
0.2069	6.0	3306	0.1243	0.9531
0.2065	7.0	3857	0.1048	0.9603
0.194	8.0	4408	0.1019	0.9636
0.1879	9.0	4959	0.0877	0.9671
0.1584	10.0	5510	0.0870	0.9687
0.1426	11.0	6061	0.0814	0.9718
0.1596	12.0	6612	0.0740	0.9749
0.1125	13.0	7163	0.0613	0.9781
0.1374	14.0	7714	0.0570	0.9787
0.1003	15.0	8265	0.0596	0.9793
0.109	16.0	8816	0.0511	0.9815
0.1206	17.0	9367	0.0497	0.9829
0.1024	18.0	9918	0.0437	0.9844
0.1051	19.0	10469	0.0420	0.9851
0.0955	20.0	11020	0.0420	0.9856

Framework versions

Transformers 4.26.1
Pytorch 1.13.1+cu116
Datasets 2.10.1
Tokenizers 0.13.2

📄 License

This model is licensed under the Apache - 2.0 license.

📦 Information Table

Property	Details
Model Type	deit-tiny-patch16-224-finetuned-main-gpu-20e-final
Training Data	imagefolder
Metrics	accuracy
License	apache-2.0

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