- get severial ImageNet pre-trained weights
- evaluate top1 and top5 accuracy on ILSVRC 2012 validation set
- Download full dataset from ILSVRC kaggle
- under Data/CLS-LOC contain following folders
- train:
- val: 50000 images
- test: 100000 images
- after unzip, there is a csv file named LOC_val_solution
- model: mobilenet_v2 trained from keras application
python mnv2_evaluation.py