Closed lauraset closed 2 years ago
Hi,
As already mentioned in the github description "Please note the method is not an exact replication of the abovementioned paper. The original code was implemented in Matlab and is not maintained/distributed anymore."
The network structure in this version can be found in networksForFeatureExtraction.py
Original code in the paper was based on a Matlab/MatConvNet, which I stopped using just a while after the publication of the original paper. This network was used in that: https://sites.google.com/site/michelevolpiresearch/codes/dense-labeling
DCVA is a general framework which can be adopted for different architectures with a little effort. The layers for feature extraction would change when a different network is used.
Hope it clarifies your question.
Hi, @sudipansaha . Thank you very much. I got it.
Hello, @sudipansaha . Your method shows promising results and inspired me a lot. However, the layer combination (2,5,8) used in your code seems different from the setting (28,16,11,6) in your paper. Can you share the network structure in this code? Thank you very much.