Models are stored depending on the way they have been trained :
ST_pretrainedmodel_public_datasetfolder contains models trained by ST using public datasetsST_pretrainedmodel_custom_datasetfolder contains models trained by ST using custom datasetsPublic_pretrainedmodel_public_datasetfolder contains public models using public datasets
Following is the overview of all pretrained image classification models available for STM32 boards. Each model family links to its folder or README for downloads, usage, and performance metrics.
Model performance analysis (TF) of these models can be used to select the model based on user's performance requirements.
Model performance analysis (Pytorch) of these models can be used to select the model based on user's performance requirements.
Note: Some folders may contain multiple model variants (Float / Int8, different input resolutions, etc.). For detailed performance tables and ONNX links, refer to the individual README of each model family.