fwtan / who_where Goto Github PK

你好，预训练模型失效了，能不能提供一下百度云链接？十分感谢

Hi sir, I get an error when demo，could you give me some advice, error as follow, and the coco.py need "val2014_filtered_persons.txt", what it contains? Thank you very much.
Traceback (most recent call last):
File "tools/demo.py", line 54, in
retriever.sample(testdb, val_imdb.objdb, mode=0, K=1, show_gt=False)
File "/mnt/hdd1/zzk/image composition/who_where-master/tools/../lib/models/RetrievalCNN.py", line 83, in sample
dst_tree = self.build_search_tree(dst_ctxdb, mode)
File "/mnt/hdd1/zzk/image composition/who_where-master/tools/../lib/models/RetrievalCNN.py", line 72, in build_search_tree
return BallTree(X, leaf_size=30, metric=cus_distance)
File "sklearn/neighbors/binary_tree.pxi", line 1068, in sklearn.neighbors.ball_tree.BinaryTree.init
ValueError: X is an empty array

Thanks for your insightful work and helpful codes! the idea is really inspiring and the codes are nicely organized.

I'm trying to train my own data on your bbox-prediction network (in PredictionCNN.py). There're 2 main issues about it:

1. Would you please kindly provide an instruction on training? (i.e. data formats, caveats, training time, etc.)
2. Variable cfg.TOP_K in function generator is undefined. I could not find any explanation about it either in the paper or in other parts of the codes.

Thanks in advance!

this is a great work, and i want to study and train the model, so can you provide the training data.thanks!

Thanks for your wonderful work and helpful codes!

I'm trying to train my own data on your bbox-prediction network (in PredictionCNN.py). I have processed the data according to what you wrote. However, i find there is no training code.It is a little hard for me to write the appropriate training codes. Could you please provide the complete codes and Instructions on training?

Your favorable reply would be highly appreciated.