Work

ConvNeXt V2-Based Dog Breed Classification: Progressive Transfer Learning on Stanford Dogs Dataset

Year: 2025

Type: article

Abstract: This repository presents a comprehensive implementation of dog breed classification using ConvNeXt V2 architecture with progressive transfer learning. Achieving 92.45% accuracy on the Stanford Dogs Dataset (120 breeds, 20,580 images), the model employs four-stage progressive unfreezing, focal loss, and advanced augmentation techniques (MixUp, CutMix). Training on dual T4 GPUs with 88M parameters. Includes complete training code, logs (26 epochs), model checkpoints, and detailed documentation.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.17597067

Author: Nasir, Huzaifa

Language: English

Topic: Advanced Neural Network Applications

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7105666263

FWCI:

Cites:

Cited by:

Related to: