Could we achieve real-time running in the Cityscapes dataset
John1231983 opened this issue · 1 comments
John1231983 commented
I have scheduled to change from CAFFE to Tensorflow. I achieve a real-time running in City-scapes data set by using CAFFE. I would like to know is it possible to run realtime using Tensorflow? It may depend on network architecture, but just assume it is FCN for simply. Thanks
fastlater commented
@John1231983 Depending on complexity of the input image, the results may vary. I tested this code and also deeplab resnet. FCN is more simple and the code is easier to read, however, the result is not as good as deeplab resnet. The problem with the second one is, the tensorflow version of the code is not well documented, the forum is closed and the code is harder to understand (at least for beginners on python and tf).