MuGeminorum/alexnet-hep2

The AlexNet-HEp2 model is specifically designed for the task of cell image classification, utilizing deep learning techniques. The training data for this model is sourced from the HEp-2 cell image dataset, which originated from the Cell Image Classification competition at the 2014 International Conference on Pattern Recognition. The dataset comprises images pre-divided into a training set (8,701 images), a validation set (2,175 images), and a test set (2,720 images). Additionally, a corresponding .csv file is provided, containing two columns: the first column consists of image IDs matching the image names in the three datasets, while the second column denotes the cell image categories. Inspired by the classical architecture of AlexNet, this model is based on a deep convolutional neural network, incorporating components such as convolutional layers, pooling layers, and fully connected layers, imparting robust capabilities for learning image features. The primary training objective is to acquire discriminative features for the HEp-2 cell image classification task, aiming to enhance classification performance on the validation and test sets. The training of this model is tailored to efficiently capture crucial information within HEp-2 cell images, facilitating accurate image classification.

Maintenance

GIT_LFS_SKIP_SMUDGE=1 git clone git@hf.co:MuGeminorum/alexnet-hep2

Training Curves

Loss	Accuracy

Mirror

https://www.modelscope.cn/models/MuGeminorum/HEp2

Reference

[1] https://github.com/MuGeminorum/Medical_Image_Computing/tree/hep2