Submit report for S1erHoR5t7 (#10)

meteor-S1erHoR5t7/README.md

@@ -0,0 +1,56 @@
+## This is a guide for [ICLR2019 reproducibility challenge](https://reproducibility-challenge.github.io/iclr_2019/)
+
+Review target: [The relativistic discriminator: a key element missing from standard GAN](https://openreview.net/forum?id=S1erHoR5t7) (*Accept Poster*)
+
+Issue Ticket: [S1erHoR5t7](https://github.com/reproducibility-challenge/iclr_2019/issues/10)
+
+Code Repo: [VSR](https://github.com/LoSealL/VideoSuperResolution)
+
+### How to reproduce benchmark
+
+0. Clone repo:
+
+    ```bash
+    git clone https://github.com/LoSealL/VideoSuperResolution -b iclr_submit
+    ```
+
+1. Download dataset and weights
+
+   1. CIFAR10 will be downloaded in the code automatically.
+   2. CelebA is downloaded [here](http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html), and is processed by:
+      ```bash
+      python VSR/Tools/DataProcessing/CelebA.py /mnt/data/dataset/celeba/ --n_test=10000
+      ```
+      where `/mnt/data/dataset/celeba` is your root to CelebA dataset. This will create a `resize64` folder holds `192,599` training patches and a `test64` folder holds `10000` testing patches under that root directory.
+
+   3. Weights are downloaded by using `python prepare_data.py --filter="\w*gan"`, which will extract weights into `./Results/` (*Requires your permission*).
+
+2. Evaluate GAN models
+
+   Example for RGAN:
+   ```bash
+   cd Train
+   python run.py --mode=eval --model=rgan --checkpoint_dir=../Results/rgan --epochs=500 --test=cifar10 --enable_inception_score --enable_fid
+   ```
+
+   Where you can see printed FID and IS value on terminal, and you can also find records file in `/tmp/vsr/<date>/eval_results.csv`.
+
+3. Generate samples
+
+   ```bash
+   cd Train
+   python run.py --model=rgan --test=cifar10
+   ```
+
+   Where the generated images are saved in `../Results/rgan/cifar10`.
+
+4. Train models from scratch
+
+    1. Refer to general guide [here](./README.md)
+    2. (Optional) Prepare your own dataset (if needed, refer DDF [here](./Data/README.md))
+    3. (Optional) Modify [model config file](./Train/parameters/rgan.yaml), all models and information are defined [here](./VSR/Models/Gan.py)
+    4. Run script: (i.e. RGAN)
+    ```bash
+    cd Train
+    python run.py --model=rgan --epochs=500 --dataset=cifar10
+    ```