initial commit

.gitignore

@@ -0,0 +1,4 @@
+*.pyc
+.ipynb_checkpoints/
+*.h5
+checkpoints/

README.md

@@ -0,0 +1,81 @@
+# Introduction
+
+This repository contains code to reproduce the experiments in [Dynamic Filter Networks](https://arxiv.org/pdf/1605.09673v2.pdf), a NIPS 2016 paper by Bert De Brabandere\*, Xu Jia\*, Tinne Tuytelaars and Luc Van Gool (\* Bert and Xu contributed equally).
+
+In a traditional convolutional layer, the learned filters stay fixed after training. In contrast, we introduce a new framework, the Dynamic Filter Network, where filters are generated dynamically conditioned on an input.
+
+Example:
+
+![mnist prediction](https://i.imgur.com/XbyD2ix.png)
+
+<!---
+![mnist gif1](https://i.imgur.com/vmkSn0k.gif)
+![mnist gif2](https://i.imgur.com/JzGhE31.gif)
+-->
+
+If you use our code in your research, please cite following paper:
+```
+@inproceedings{debrabandere16_dfn,
+  author    = {Bert De Brabandere and Xu Jia and Tinne Tuytelaars and Luc Van Gool},
+  title     = {Dynamic Filter Networks},
+  booktitle = {NIPS},
+  year      = {2016}
+}
+```
+
+# Running the code
+
+* Install [Lasagne](https://lasagne.readthedocs.io/en/latest/user/installation.html) and its prerequisites.
+* Download the datasets and update the paths in the datasets/dataset_*.py files to point to them:
+```
+wget http://www.cs.toronto.edu/~emansim/datasets/mnist.h5
+wget https://homes.esat.kuleuven.be/~bdebraba/dfn/movingObjects.h5
+wget https://homes.esat.kuleuven.be/~bdebraba/dfn/highwayDriving_train.h5
+wget https://homes.esat.kuleuven.be/~bdebraba/dfn/highwayDriving_test.h5
+```
+
+* Run the experiments:
+```
+python experiment_steerableFilter.py
+python experiment_bouncingMnistOriginal.py
+python experiment_highwayDriving.py
+python experiment_stereoPrediction.py
+```
+This will write checkpoint files to the checkpoints directory.
+
+* You can also run the baseline models. They have the same architecture as the DFN models, but without the DFN layer at the end:
+```
+python experiment_bouncingMnistOriginal_baseline.py
+python experiment_highwayDriving_baseline.py
+python experiment_stereoPrediction_baseline.py
+```
+Finally, you can evaluate the DFN and baseline models on the test set and generate new predictions with the notebook files:
+```
+analyse_trained.ipynb
+analyse_trained-baseline.ipynb
+```
+
+## Tensorflow implementation
+While we don't provide a full implementation of the experiments in tensorflow, an example dynamic filter layer can be found in `layers/dynamic_filter_layer_tensorflow.py`.
+
+# Results
+
+When evaluating the trained models on the test sets with the ipython notebooks, you should approximately get following results:
+
+| Loss (per pixel)      | Baseline | DFN       |
+| --------------------- |:--------:| ---------:|
+| Moving MNIST (bce)    | 0.106144 | 0.068914  |
+| Highway Driving (mse) | 0.003683 | 0.003270  |
+| Stereo Cars (mse)     | 0.000416 | 0.000330  |
+
+| Loss (image, 64x64)   | Baseline | DFN      |
+| --------------------- |:--------:| --------:|
+| Moving MNIST (bce)    | 434.8    | 282.3    |
+| Highway Driving (mse) | 15.08    | 13.39    |
+| Stereo Cars (mse)     | 1.70     | 1.35     |
+
+| # Params        | Baseline  | DFN     |
+| --------------- |:---------:|:-------:|
+| Moving MNIST    | 637,443   | 637,361 |
+| Highway Driving | 368,245   | 368,122 |
+| Stereo Cars     | 464,509   | 464,494 |