AIbase
EN
Home
10E Affecthq Fer Balanced W0.1 Jitter Jiggle
1

10E Affecthq Fer Balanced W0.1 Jitter Jiggle

Developed by Piro17
A vision model fine-tuned from google/vit-base-patch16-224-in21k for facial expression recognition tasks
Face-related
Transformers
Open Source License:Apache-2.0#Fine-tuned ViT model#Emotion recognition#Facial expression analysis
Downloads 17
Release Time : 2/15/2023

Model Overview

This model is a facial expression recognition model based on the ViT architecture. After fine-tuning on an unknown dataset, it achieved approximately 60% accuracy on the evaluation set.

Model Features

Balanced training
Utilized balanced training strategies, possibly employing class weights or data augmentation techniques
Data augmentation
Used data augmentation techniques such as jitter and jiggle during training
Stable training
Through 10 epochs of training, the model's performance improved steadily without overfitting

Model Capabilities

Facial expression recognition
Image classification

Use Cases

Emotion analysis
Real-time expression recognition
Can be used for real-time expression analysis in video conferences or social media
Accuracy approximately 60%
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
EnglishįŽ€äŊ“中文įšéĢ”中文ãĢãģんご
Š 2025AIbase