heefe92
Problem with experimental results
I ran the relation code based on the vgg16 network, but found that the number of epochs is the same, but much lower than the map value of the individual faster-rcnn code. I don't understand why this is the case. Shouldn't the map value be increased in the paper? Is this the case when you do the experiment? I want to ask you, thank you.
I think the code is still uncompleted. Because the Relation-Module was not added in the FasterRCNN. You can see the line 66 in model.py, n_realtions = 0. And when I change it to 2, the code cannot run more. Do u notice this problem?
is the code still uncompleted? I'm a novice on deeplearning, I need an implementation of RN based on pytorch.Can somebody tell me whether the author is still following this code? thanks
@puyiwen Hello, i have few problems about the 'RelationModule'. What do the parameters, 'appearance_feature_dim', 'key_feature_dim', 'geo_feature_dim' mean? Thanks very much
