Commit 7115e191 authored by Isaak Lim's avatar Isaak Lim

edit README.md

parent 76a3cfe1
......@@ -19,17 +19,20 @@ We used our own sampling of the [Shapenet][shapenet] Dataset, which can be downl
To evaluate a model run
python test_autoencoder.py --model-path pretrained/model_full_transformations.state --data-path "data/shapenet_v2_16000.hdf5"
--split-path "data/core_train_val_test.csv"
python test_autoencoder.py --model-path pretrained/model_full_transformations.state \
--data-path "data/shapenet_v2_16000.hdf5" \
--split-path "data/core_train_val_test.csv"
pretrained weights for our models trained with 3 or 9 layers of adaptive instance normalization can be found under models/pretrained/
When evaluating the version with only 3 used layers add the option --adain-layer 3
to train a new model run
Pretrained weights for our models trained with 3 or 9 layers of adaptive instance normalization can be found under models/pretrained/.
python train_autoencoder.py --run-name full --data-path "data/shapenet_v2_16000.hdf5"
--split-path "data/core_train_val_test.csv"
When evaluating the version with only 3 used layers add the option --adain-layer 3 to train a new model run
python train_autoencoder.py --run-name full \
--data-path "data/shapenet_v2_16000.hdf5" \
--split-path "data/core_train_val_test.csv"
The weights will be saved under models/, whereas the progress will be saved under runs/.
the weights will be saved under models/, whereas the progress will be saved under runs/
There are further command line options to specify the used losses etc.
[shapenet]: https://www.shapenet.org/
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment