process_dataset in extract_cnn_features -- is it a bug?

[micklexqg]

Hi, I believe there is a bug in the following function: ` def process_dataset(dataset, net, gpu_id): data_dir = paths.dataset_dir[dataset] + '/' images_dir = paths.images_dir[dataset] data = json.load(open(data_dir + 'dataset_%s.json' % dataset, 'r'))

splits = defaultdict(list)
for im in data['images']:
    split = im['split']
    if split == 'restval':
        split = 'train'
    splits[split].append(images_dir[dataset] + im['filepath'] + '/' + im['filename'])

`

I evaluate process_dataset('coco', 'VGG19', 1), and "string indices must be integers, not str" error appear at the line "splits[split].append(images_dir[dataset] + im['filepath'] + '/' + im['filename'])". I debug and find maybe it has something wrong with images_dir[dataset]. @ivendrov, is it a bug?