Skip to content

Latest commit

 

History

History
21 lines (19 loc) · 921 Bytes

README.md

File metadata and controls

21 lines (19 loc) · 921 Bytes

multi_gpu_image_feature_extractor

pytorch;multi gpu;image feature;

This is an implementation of extractig feature from image with resnet(pretrained on imagenet or place365) or swin_transformer.

run

Download the resnet paras pretrained in imagenet resnet_imagenet (psw:ptal)or resnet paras pretrained in place365 resnet_place365 (psw:i0d8) or swin_transformer_l_224 at swin_transformer run the code by:

cd multi_gpu_image_feature_extractor
sh extract.sh

or:

python extrat_image_feature_from_image.py \
    --imagelist_file your_image_file_list \
    --out_path_root your_featurefile_save_path \
    --modelname resnet_imagenet \
    -b 48

you can set the modelname as "resnet_imagenet", "resnet_place365" or "swin_base_patch4_window12_384" so far