hq_fer2013 Open - source Facial Expression Recognition Model - Accurately recognize facial expressions with an accuracy rate of up to 70.22%

Hq Fer2013

Developed by Piro17

A facial expression recognition model fine-tuned based on Google's ViT model, trained on the FER2013 dataset with an accuracy of 70.22%.

Image Classification

Transformers

Open Source License:Apache-2.0 #ViT fine-tuned model #Image classification #Facial expression recognition

Downloads 38

Release Time : 2/17/2023

Model Overview

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on an image folder dataset, primarily used for facial expression recognition tasks.

Model Features

High accuracy

Achieves 70.22% accuracy on the FER2013 dataset, outperforming the baseline model.

Based on ViT architecture

Utilizes the advanced Vision Transformer architecture to effectively capture global image features.

Fine-tuned

Undergone 13 rounds of fine-tuning with stable improvements across all metrics.

Model Capabilities

Facial expression recognition

Image classification

Sentiment analysis

Use Cases

Affective computing

Real-time expression recognition

Used for real-time expression analysis in video conferences or social media

Can recognize 7 basic expressions with 70.22% accuracy

User experience research

Analyze users' emotional responses to products or content

🚀 hq_fer2013

This model is a fine - tuned version of google/vit-base-patch16-224-in21k on the imagefolder dataset. It can be used for image classification tasks, providing reliable accuracy, precision, recall, and F1 scores.

🚀 Quick Start

This model is a fine - tuned version of google/vit-base-patch16-224-in21k on the imagefolder dataset. It achieves the following results on the evaluation set:

Loss: 0.8438
Accuracy: 0.7022
Precision: 0.7039
Recall: 0.7022
F1: 0.7022

📚 Documentation

Model Information

Property	Details
Model Type	Fine - tuned version of google/vit-base-patch16-224-in21k
Training Data	imagefolder dataset
Metrics	Accuracy, Precision, Recall, F1

Training Procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e - 05
train_batch_size: 32
eval_batch_size: 32
seed: 17
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: 13

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
1.3081	1.0	398	1.3132	0.5555	0.5079	0.5555	0.5137
0.991	2.0	796	1.0141	0.6332	0.6356	0.6332	0.6153
0.9099	3.0	1194	0.9257	0.6682	0.6677	0.6682	0.6631
0.8306	4.0	1592	0.8832	0.6765	0.6838	0.6765	0.6747
0.7755	5.0	1990	0.8583	0.6892	0.6896	0.6892	0.6876
0.7129	6.0	2388	0.8442	0.6931	0.6951	0.6931	0.6922
0.6549	7.0	2786	0.8494	0.6952	0.7054	0.6952	0.6978
0.6246	8.0	3184	0.8394	0.6963	0.7023	0.6963	0.6977
0.6138	9.0	3582	0.8421	0.6996	0.7080	0.6996	0.7013
0.5824	10.0	3980	0.8438	0.7022	0.7039	0.7022	0.7022
0.5517	11.0	4378	0.8497	0.7002	0.7034	0.7002	0.7005
0.5154	12.0	4776	0.8508	0.7021	0.7030	0.7021	0.7018
0.5318	13.0	5174	0.8540	0.7010	0.7029	0.7010	0.7013

Framework versions

Transformers 4.27.0.dev0
Pytorch 1.13.1+cu116
Datasets 2.9.0
Tokenizers 0.13.2

📄 License

This model is licensed 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