Train and Read Charades Dataset, implemented on Python and PyTorch, on Windows.
Based on PyTorch Starter Code for Activity Classification and Localization on Charades
You have to download trained files.
> bash
/charades-algorithms$ ./get_alreadytrained.sh
You have to download csv files from Charades Data Set and locate it on ./ folder.
You have to fix routes to datasets folder in datasets/charadesrgb.py and datasets/charadesflow.py.
Show specific video!
charades-algorithms> python read.py -id={VIDEO_ID}
Save specific video!
Save it at
D:/workspace-dataset/charades/Charades_v1_with_actions. You have to fix location inread.py.
charades-algorithms> python read.py -id={VIDEO_ID} -s=True
Save all video!
charades-algorithms> python read.py -id=all
Train RGB!
charades-algorithms> python exp/rgbnet.py --data={PATH_TO_DATASET_FOLDER}
Train FLOW!
charades-algorithms> python exp/flownet.py --data={PATH_TO_DATASET_FOLDER}
Test with pretrained-model!
charades-algorithms> python exp/rgbnet.py --pretrained-weights={PATH_TO_PRETRAINED_WEIGHTS_FILE-~.pth.tar}