Merge pull request #2 from hslu-abiz/refactoring

Finished Refactoring

.dockerignore

@@ -0,0 +1,9 @@
+data
+data/*
+data/*/*
+tmp
+tmp/*
+profiler
+profiler/*
+profiler_env/*
+core

.gitattributes

@@ -0,0 +1,4 @@
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.pdf filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -1,4 +1,3 @@
-
 # Created by https://www.toptal.com/developers/gitignore/api/vim,python,pycharm+all,jupyternotebooks
 # Edit at https://www.toptal.com/developers/gitignore?templates=vim,python,pycharm+all,jupyternotebooks
 
@@ -277,3 +276,5 @@ extended_prototype/saved_weights/*
 extended_prototype/soundbank/*
 extended_prototype/soundscapes/*
 extended_prototype/test_segmentations/*
+saved_weights/*
+experiments/*

Dockerfile

@@ -0,0 +1,21 @@
+FROM tensorflow/tensorflow:2.4.0-gpu
+
+RUN apt-get update && apt-get install -y apt-transport-https
+RUN apt-get install -y libtcmalloc-minimal4
+RUN apt-get install -y sox
+
+RUN apt install -y libsndfile1
+RUN apt install -y libsm6 libxext6 libxrender-dev
+
+RUN pip install --upgrade pip
+
+WORKDIR /tf
+
+RUN mkdir /assets
+
+COPY requirements.txt /assets/requirements.txt
+RUN pip install -r /assets/requirements.txt --upgrade --no-cache-dir
+
+COPY . /tf/
+
+RUN ./scripts/install.sh

README.md

@@ -1,5 +1,7 @@
 # Soundscape Generation
 
+Generate soundscapes from images.
+
 ## Table of Contents
 
 1. [Installation](#installation)
@@ -17,7 +19,7 @@ Follow the instructions give in the following link:
 
 * [Scaper installation](https://scaper.readthedocs.io/en/latest/installation.html)
 
-### Download Dependencies
+### Install Dependencies
 
 ```bash
 pip install -r requirements.txt
@@ -30,7 +32,7 @@ on [www.cityscapes-dataset.com](https://www.cityscapes-dataset.com/). After the
 script. During the download, it will ask you to provide your email and password for authentification.
 
 ```bash
-./download_data.sh
+./scripts/download_data.sh
 ```
 
 ## Usage
@@ -41,31 +43,35 @@ finetuned on the Cityscapes dataset.
 
 ### Train Object Segmentation Network
 
-To train the network, run the follwing command.
+To train the network, run the follwing command. The hyperparameters epoch and batch size can be configured in the `docker-compose.yml` file. To load a pre-trained model specify its path in the `MODEL_TO_LOAD` variable, if the variable is `None` the model is trained from scratch.
 
 ```bash
-python train.py --num_epochs 70 --batch_size 8 --evaluate_every 1 --save_weights_every 1
+docker-compose up train_object_detection
 ```
 
-By default, training resumes from the latest saved checkpoint. If the `checkpoints/` directory is missing, the training
-starts from scratch.
-
 ### Test the Segmentation Network
 
-Run the following command to predict the semantic segmentation of every image in the `test_images/` directory (note:
-results are saved in the `test_segmentations/` directory)
+Run the following command to predict the semantic segmentation of every image in the `--test_images` directory (note:
+predictions are saved with the same name and a `_pred.jpg` suffix). Ensure that you specify the correct image's file type in `--test_images_type`.
 
 ```bash
-python predict.py
+docker-compose up predict_object_detection
 ```
 
-Ensure that you specify the image's file type in the image path variable in `predict.py`.
+### Evaluate the Segmentation Network
+To evaluate the segmentation network run the command below.
+
+```bash
+docker-compose up evaluation
+```
 
 ### Generate soundscapes
 
-Run the file soundGeneration.py to generate soundscapes of every image in the `test_images/` directory (note: results
-are saved in the `soundscapes/` directory). Ensure that you specify the image type of the image in the image path
-variable of `predict.py`.
+To generate soundscapes of every image in the `--test_images` directory run the following command. The generated audios will be saved in `data/soundscapes`. Ensure that you specify the correct image's file type in `--test_images_type`.
+
+```bash
+docker-compose up sound_generation
+```
 
 ## Results
 

docker-compose.yml

@@ -0,0 +1,71 @@
+version: "3.2"
+services:
+    train_object_detection:
+        build: .
+        volumes:
+            - ${PWD}:/tf/
+        working_dir: /tf
+        command: bash -c "./scripts/train_object_detection.sh"
+        devices:
+            - /dev/nvidia0
+        environment:
+            NVIDIA_VISIBLE_DEVICES: 0
+            EPOCHS: 70
+            BATCH_SIZE: 8
+            MODEL_TO_LOAD: None
+        deploy:
+            resources:
+                reservations:
+                    devices:
+                        - capabilities: [ gpu ]
+
+    predict_object_detection:
+        build: .
+        volumes:
+            - ${PWD}:/tf/
+        working_dir: /tf
+        command: bash -c "./scripts/predict_object_detection.sh"
+        devices:
+            - /dev/nvidia0
+        environment:
+            NVIDIA_VISIBLE_DEVICES: 0
+            WEIGHTS_PATH: None
+        deploy:
+            resources:
+                reservations:
+                    devices:
+                        - capabilities: [ gpu ]
+
+    sound_generation:
+        build: .
+        volumes:
+            - ${PWD}:/tf/
+        working_dir: /tf
+        command: bash -c "./scripts/sound_generation.sh"
+        devices:
+            - /dev/nvidia0
+        environment:
+            NVIDIA_VISIBLE_DEVICES: 0
+            WEIGHTS_PATH: None
+        deploy:
+            resources:
+                reservations:
+                    devices:
+                        - capabilities: [ gpu ]
+
+    evaluation:
+        build: .
+        volumes:
+            - ${PWD}:/tf/
+        working_dir: /tf
+        command: bash -c "./scripts/evaluation.sh"
+        devices:
+            - /dev/nvidia0
+        environment:
+            NVIDIA_VISIBLE_DEVICES: 0
+            WEIGHTS_PATH: None
+        deploy:
+            resources:
+                reservations:
+                    devices:
+                        - capabilities: [ gpu ]

requirements.txt

@@ -1,15 +1,13 @@
-python==3.8.0
-numpy==1.19.5
-pandas==1.2.4
-scikit-learn==0.24.2
-matplotlib==3.4.2
-scipy==1.6.3 
-sox==1.4.0
-ffmpeg==4.3.1
-ffmpeg-python==0.2.0
+numpy
+pandas
+scikit-learn
+matplotlib
+scipy
+Pillow
+sox
+ffmpeg
+ffmpeg-python
 scaper==1.6.5
-tensorflow==2.5.0
-cudatoolkit==11.0.221
-tensorflow-addons==0.13.0
-pillow==7.1.2
-cityscapesscripts==2.2.0
+tensorflow>=2.5.0
+tensorflow-addons>=0.13.0
+cityscapesscripts==2.2.0

scripts/evaluation.sh

@@ -0,0 +1 @@
+LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4 python -m soundscape_generation.evaluation --weights $WEIGHTS_PATH

scripts/install.sh

@@ -0,0 +1,10 @@
+apt-get update
+
+apt-get -y install git
+apt-get -y install libav-tools
+apt-get -y install libsndfile1-dev
+apt-get -y install libsndfile1
+apt-get -y install libcupti-dev
+
+pip install --upgrade pip
+pip install -r requirements.txt

scripts/predict_object_detection.sh

@@ -0,0 +1 @@
+LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4 python -m soundscape_generation.predict --weights $WEIGHTS_PATH

scripts/sound_generation.sh

@@ -0,0 +1 @@
+LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4 python -m soundscape_generation.sound_generation --weights $WEIGHTS_PATH

scripts/train_object_detection.sh

@@ -0,0 +1 @@
+LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4 python -m soundscape_generation.train --num_epochs $EPOCHS --batch_size $BATCH_SIZE --model_to_load $MODEL_TO_LOAD

setup.py

@@ -29,8 +29,8 @@ def parse_requirements(filename):
 
 README_MD = open(join(dirname(abspath(__file__)), "README.md")).read()
 
-setup(name='soundscape-generation',
-      version=get_version('soundscape-generation/__init__.py'),
+setup(name='soundscape_generation',
+      version=get_version('soundscape_generation/__init__.py'),
       author='ABIZ Lab',
       author_email='[email protected]',
       description='Generate soundscapes based on images.',