dhruvsharma15/medvqa

medvqa's Issues

Data Folder NOT available

There is supposed to be a Data directory which probably has the pretrained models e.g. ../Data\resnet152\train_fc7.h5.

But this is NOT available in the files and folders present in the repository and hence, we cannot run the code.

Usage:

def load_image_features(data_dir='../Data', model='resnet152', split='Train'):
    fc7_features = None
    image_id_list = None
    with h5py.File( join( join(data_dir, model), (split + '_fc7.h5')),'r') as hf:
        fc7_features = np.array(hf.get('fc7_features'))
    with h5py.File( join( join(data_dir, model), (split + '_image_id_list.h5')),'r') as hf:
        image_id_list = np.array(hf.get('image_id_list'))
        image_id_list = [id.decode('UTF-8') for id in image_id_list]
        
    return fc7_features, image_id_list

def load_ques_embed(data_dir='../Data', model='bert', split='train', cat='allQA'):
    embeddings = None
    image_id_list = None
    cat_name = ""
    if(cat=='allQA'):
        cat_name = '_All_QA_Pairs_'
    elif(cat=='cat1'):
        cat_name = '_C1_Modality_'
    elif(cat=='cat2'):
        cat_name = '_C2_Plane_'
    elif(cat=='cat3'):
        cat_name = '_C3_Organ_'
    else:
        cat_name = '_C4_Abnormality_'
        
    root_path = join(join(join(data_dir, model), cat))
    with h5py.File( join( root_path, (split + cat_name +split+'_embed.h5')),'r') as hf:
        embeddings = np.array(hf.get('embedding'))
    with h5py.File( join( root_path, (split + cat_name +split+'_image_id_list.h5')),'r') as hf:
        image_id_list = np.array(hf.get('image_id_list'))
        image_id_list = [id.decode('UTF-8') for id in image_id_list]
        
    return embeddings, image_id_list

Pls. update the code by providing the Data directory or if you could share a drive link for the above said folder.

@dhruvsharma15