Instructions are clear, dependencies were easily installed with pip. I only ran into a CudNN issue: "E tensorflow/stream_executor/cuda/cuda_dnn.cc:319] Loaded runtime CuDNN library: 7.5.0 but source was compiled with: 7.6.0." which was easily fixed by reinstalling CudNN. The code produces many warnings, during initialization (e.g., "WARNING:tensorflow:From test_vectorization.py:325: The name tf.global_variables_initializer is deprecated. Please use tf.compat.v1.global_variables_initializer instead." or "WARNING:tensorflow:From NeuralRenderer.py:69: The name tf.depth_to_space is deprecated. Please use tf.compat.v1.depth_to_space instead" or "WARNING:tensorflow:From VGG16.py:80: max_pooling2d (from tensorflow.python.layers.pooling) is deprecated and will be removed in a future version. Instructions for updating: Use keras.layers.MaxPooling2D instead.") and runtime (e.g., "test_vectorization.py:74: RuntimeWarning: invalid value encountered in true_divide" or "W tensorflow/core/common_runtime/colocation_graph.cc:983] Failed to place the graph without changing the devices of some resources. Some of the operations (that had to be colocated with resource generating operations) are not supported on the resources' devices. Current candidate devices are [ /job:localhost/replica:0/task:0/device:CPU:0]."), but nothing that prevents from getting reasonable outputs. The deprecated functions might indicate that the code may not run in the near future. A small difference with the paper is acknowledged : the stroke thickness in the exported svg files is constant, while it is varying in the paper. While it precludes direct and exact comparisons, the effect of the varying stroke thickness in the paper is very subtle.
If you want to contribute with another review, please follow these instructions.
Please consider to cut/paste/edit the raw JSON data attached to this paper.