ResNet18 CIFAR-10 Open-Source Image Classification Model - Trained on CIFAR-10 with Nearly 95% Testing Accuracy

Resnet18 Cifar10

Developed by edadaltocg

A compact ResNet18 model trained on the CIFAR-10 dataset, achieving a test accuracy of 94.98%.

Image Classification EnglishOpen Source License:MIT #Lightweight Image Classification #High Accuracy on CIFAR10 #Fast Inference

Downloads 151

Release Time : 2/19/2023

Model Overview

This model is a lightweight ResNet18 architecture, optimized for CIFAR-10 image classification tasks, suitable for fast image classification scenarios.

Model Features

High Accuracy

Achieves a classification accuracy of 94.98% on the CIFAR-10 test set.

Lightweight Architecture

Utilizes the ResNet18 architecture with fewer parameters, making it suitable for resource-constrained environments.

Fast Inference

The lightweight model optimized for CIFAR-10 data enables rapid image classification.

Model Capabilities

Image Classification

10-Class Recognition

Use Cases

Education & Research

Image Classification Teaching

Used as an example in introductory deep learning courses for image classification.

Helps students understand the fundamental principles of convolutional neural networks.

Prototype Development

Rapid Proof of Concept

Provides quick prototyping capabilities for image recognition applications.

Validates the feasibility of image classification features in a short time.

Property	Details
Model Type	resnet18_cifar10
Training Data	CIFAR - 10
Testing Data	CIFAR - 10
Test Accuracy	0.9498
License	MIT

Parameter	Value
config	`scripts/train_configs/cifar10.json`
model	`resnet18_cifar10`
dataset	`cifar10`
batch_size	`128`
epochs	`300`
validation_frequency	`5`
seed	`1`
criterion	`CrossEntropyLoss`
criterion_kwargs	`{}`
optimizer	`SGD`
lr	`0.1`
optimizer_kwargs	`{'momentum': 0.9, 'weight_decay': 0.0005, 'nesterov': 'True'}`
scheduler	`ReduceLROnPlateau`
scheduler_kwargs	`{'factor': 0.1, 'patience': 3, 'threshold': 0.001, 'mode': 'max'}`
debug	`False`