Updated README.md

README.md

@@ -4,13 +4,23 @@ Implemented Architecture of pyramid scene parsing network in Keras
 
 Converted trained weights needed to run the network.
 
+
 Download converted weights here:
 [link:pspnet.npy](https://www.dropbox.com/s/9xebhix7dbk372d/pspnet.npy?dl=0)
 
 And place in directory with pspnet.py
 
+Weights from caffemodel were converted with [caffe-tesnorflow](https://github.com/ethereon/caffe-tensorflow), source code of coverter was modified to fit batch normalization, which is annotated as 'BN' in original prototxt
+
+Interpolation layer is implemented in code as custom layer
+
+## Important
+
+This implementation is not working properly despite calculations are made without errors(output image is very very bad).
+I can't figure out which causes such behavior, so help and proposals are appreciated.
+
 Memory usage:3500Mb
-Calculation speed: 1.2 sec
+Calculation speed: 1.2 sec on gtx 1080
 
 ## Dependencies:
 1. Tensorflow
@@ -22,4 +32,5 @@ Calculation speed: 1.2 sec
 
 ```bash
 python pspnet.py
-```
+```
+this causes to start