I have trained model with PyTorch on the Imagenet,Places365 and so on.
Model | Dataset | top1 | top5 |
---|---|---|---|
resnet50 | Imagenet | 76.364 | 93.122 |
senet50 | Imagenet | 77.264 | 93.612 |
resnet50 | Places365 | 55.18 | 85.29 |
densenet161 | Places365 | - | - |
python load_bestmodel.py [model_path]
python main.py -a resnet50 -e [data-folder with train and val folders]
if you use places365 dataset, you should change the code of data_load, and change the number of classes