lantuzi
lantuzi
gt_pascal.pkl is the training data, each row is one object. Its structure is same as and created by PASCAL_VOC/get_data_from_XML.py as I found. Thanks to the great work of rykov8
I have the same problem with changing image size, no matter which prior_box to use. Could please help and give hint @rykov8 . I think it is related to number...
Hi, @oarriaga I still use the code from repo with slight change. So the training has this one in 'SSD_training.ipynb': yield preprocess_input(tmp_inp), tmp_targets and the test has this one in...