Closed AakashKumarNain closed 7 years ago
We are really sorry about making mistakes in our code.
This error is caused by cnn_feature_extractor.py
, which should save pool5 feature as float type but it currently save pool5 features as double type.
I will fix it soon but as a workaround, you can manually do type casting when loading pool5 feature.
e.g. In utils/data_utils.py
, line 91
np.load(os.path.join(FLAGS.img_data_dir, batch_path[i])).astype(np.float32)
Okay. Thanks!!
When I ran the
train.py
file using the options asnum_gpus 1
andbatch_size 10
, it throws the following error :