Inspired by COVID-Next and the efficiency and mobility of Efficientnet, we are now open sourcing the upgraded Pytorch implementation of both called COVID-Efficientnet.
COVID-Efficientnet features an architecture that builds upon Efficientnet b7 architecture, an AutoML architecture for optimizing both accuracy and mobility.
Code was tested with Python 3.7.
We use the following datasets: * COVID ChestXray dataset. Be aware this repository is constantly adding new images. * Kaggle RSNA pneumonia dataset
The following results were obtained on the dataset used in the original repo as of May 19 2020. | Model | Accuracy | |:—————–:|:——–:| | COVID-Net (Large) | 91.90% | | COVID-Next | 94.76% | | COVID-Efficientnet | 96.01% |
Pretrained weights are available at weights