Docker setup + commit Monodepth2

.env

@@ -0,0 +1,3 @@
+SERVICE=seathru
+WORKING_DIR=/opt/app
+PLATFORM=linux/x86_64

.gitignore

@@ -0,0 +1,4 @@
+__pycache__
+.DS_Store
+*.png
+models/

.gitmodules

@@ -1,3 +0,0 @@
-[submodule "deps/monodepth2"]
-	path = deps/monodepth2
-	url = [email protected]:nianticlabs/monodepth2.git

Dockerfile

@@ -0,0 +1,41 @@
+ARG PLATFORM
+
+FROM --platform=${PLATFORM} python:latest
+
+# These will be pulled from the docker-compose.yml file via the `args` parameter
+ARG WORKING_DIR
+ARG CONDA_ENV_NAME
+
+WORKDIR ${WORKING_DIR}
+
+# Set some known env vars + lift from docker-compose args into env
+ENV CONDA_ENV_NAME ${CONDA_ENV_NAME}
+ENV CONDA_DIR /opt/conda
+ENV PATH=$CONDA_DIR/bin:$PATH
+
+# Dependencies
+RUN apt-get update && apt-get install -y \
+    cmake \
+    build-essential \
+    wget \
+    gettext-base \
+    ffmpeg \
+    libsm6 \
+    libxext6 && \
+    apt-get clean
+
+# Install miniconda (Basically anaconda without all the defaults, just the CLI.)
+# This'll let us use whichever version of python is compatible with seathru and monodepth2. Should also make it less painful to upgrade in the future.
+RUN wget --quiet https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda.sh && \
+    /bin/bash ~/miniconda.sh -b -p $CONDA_DIR
+
+COPY conda-environment.yml ./conda-environment-template.yml
+RUN cat conda-environment-template.yml | envsubst > environment.yml
+
+RUN conda env create -f environment.yml
+RUN echo "source activate $CONDA_ENV_NAME" > ~/.bashrc
+ENV PATH $CONDA_DIR/envs/$CONDA_ENV_NAME/bin:$PATH
+
+ADD . ${WORKING_DIR}
+
+ENTRYPOINT ["tail", "-f", "/dev/null"]

README.md

@@ -0,0 +1,28 @@
+# Sea-Thru
+Implementation of Sea-thru by Derya Akkaynak and Tali Treibitz
+
+__Forked from https://github.com/hainh/sea-thru__
+
+This fork's only aim is to standardize the dependencies and environment with [Docker](https://docs.docker.com/get-docker/) and [Anaconda](https://www.anaconda.com/) to make it so that anyone can use it without deep technical knowledge of python.
+
+Additionally, in order to make this an all-in-one package, the [Monodepth2 submodule](https://github.com/nianticlabs/monodepth2/tree/b676244e5a1ca55564eb5d16ab521a48f823af31) has been cloned and commited directly into this repo, preserving the commit history and original authors.
+
+## Prerequisites
+
+[Docker](https://docs.docker.com/get-docker/)
+
+## Setup
+
+1. Run `docker compose up --build`
+    - This can take around 2 hours the first time; from then onwards you only need to run `docker compose up` to get the container running and it'll boot up instantly.
+2. Once the container is built and running, in a separate terminal run `docker exec -it seathru /bin/bash`
+    - Replace `seathru` in that command if you've modified the `SERVICE` enviornment variable in `.env`
+3. From within the container you can now run `python seathru-mono-e2e.py --image ${PATH_TO_IMAGE}`
+    - This root directory is mounted into the container each time you bring it up via `docker compose up`, meaning any files you add on your host machine to this directory will _also_ be available within the container.
+
+## Description
+
+A recent advance in underwater imaging is the Sea-Thru method, which uses a physical model of light attenuation to reconstruct
+the colors in an underwater scene. This method utilizes a known
+range map to estimate backscatter and wideband attenuation
+coefficients. This range map is generated using structure-from-motion (SFM), which requires multiple underwater images from various perspectives and long processing time. In addition, SFM gives very accurate results, which are generally not required for this method. In this work, we implement and extend Sea-Thru to take advantage of convolutional monocular depth estimation methods, specifically the Monodepth2 network. We obtain satisfactory results with the lower-quality depth estimates with some color inconsistencies using only one image.

conda-environment.yml

@@ -0,0 +1,27 @@
+# Conda Environment https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html
+name: $CONDA_ENV_NAME
+channels:
+  - conda-forge
+  - pytorch
+dependencies:
+ # Dependencies for sea-thru migrated from requirements.txt https://github.com/hainh/sea-thru/blob/master/requirements.txt
+  - scikit-learn=0.24.2
+  - numpy=1.19.2
+  - matplotlib=2.2.2
+  - pillow=5.1.0
+  - scipy=1.5.2
+  - scikit-image=0.16.2
+  - pynng=0.7.1
+  - pyyaml=5.4.1
+
+  # Lock pip version as well to install rawpy, wasn't able to install through conda/mamba
+  - pip=21.3.1
+  - pip:
+    - rawpy==0.16.0
+
+  # Dependencies and python version for Monodepth2 described here: https://github.com/nianticlabs/monodepth2#%EF%B8%8F-setup
+  - python=3.6.6
+  - pytorch=0.4.1
+  - torchvision=0.2.1
+  - tensorboardX=1.4
+  - opencv=3.3.1

deps/monodepth2/.github/ISSUE_TEMPLATE/problem-training-on-kitti.md

@@ -0,0 +1,15 @@
+---
+name: Problem training on KITTI
+about: Issue when training with default settings
+title: ''
+labels: ''
+assignees: ''
+
+---
+
+If you have problems training on KITTI, please make sure to check:
+
+1. That you are using an unmodified version of the codebase
+2. That you are using the correct versions of the dependencies we specified in the README
+
+If you’ve checked both of these, please feel report here the problem you are having, together with your hardware and anything else potentially unusual about your setup.

deps/monodepth2/.github/ISSUE_TEMPLATE/training-on-custom-training-data.md

@@ -0,0 +1,28 @@
+---
+name: Training on custom training data
+about: Problems training on data which isn't KITTI
+title: ''
+labels: custom-dataset
+assignees: ''
+
+---
+
+Thanks for trying out monodepth2 on a different dataset!
+
+Before logging your issue, please look through the issues already tagged with [custom dataset](https://github.com/nianticlabs/monodepth2/issues?q=label%3Acustom-dataset+)  , as your problem may well have been answered there.
+
+In particular please note that:
+
+1. You should know the intrinsics for your custom dataset, and you should set them *normalized* in your dataloader. More details are in the [dataloader comments](https://github.com/nianticlabs/monodepth2/blob/master/datasets/kitti_dataset.py#L24).
+
+2. KITTI is captured at 10 frames per second, and this seems to work well for training in our repo. Most videos are higher frame rate than this, so you may want to consider temporally downsampling your dataset (or setting `--frame_ids` appropriately)
+
+3. If you have moving objects in your training you are likely to still see ‘holes’ punched in your predicted depths, even with automasking turned on (https://github.com/nianticlabs/monodepth2/issues/310) 
+
+4. Monodepth2 is unlikely to work on monocular training data from indoor environments, or captured from difficult camera motions. It is best suited to driving scenarios or other simple forward camera motions.
+
+If you still have issues training, we might be able to help. Please report:
+
+1. What you have changed to create your custom dataset
+2. Some example training images
+3. What do the predicted depths in the tensorboard output look like?

deps/monodepth2/.gitignore

@@ -0,0 +1,7 @@
+__pycache__
+*.pyc
+*_disp.jpg
+*_disp.npy
+*.npz
+kitti_data
+models

deps/monodepth2/LICENSE

@@ -0,0 +1,182 @@
+Copyright © Niantic, Inc. 2018. Patent Pending.
+
+All rights reserved.
+
+
+
+================================================================================
+
+
+
+This Software is licensed under the terms of the following Monodepth2 license
+which allows for non-commercial use only. For any other use of the software not
+covered by the terms of this license, please contact [email protected]
+
+
+
+================================================================================
+
+
+
+Monodepth v2 License
+
+
+    This Agreement is made by and between the Licensor and the Licensee as
+defined and identified below.
+
+
+1.  Definitions.
+
+    In this Agreement (“the Agreement”) the following words shall have the
+following meanings:
+
+    "Authors" shall mean C. Godard, O. Mac Aodha, M. Firman, G. Brostow
+    "Licensee" Shall mean the person or organization agreeing to use the
+Software in accordance with these terms and conditions.
+    "Licensor" shall mean Niantic Inc., a company organized and existing under
+the laws of Delaware, whose principal place of business is at 1 Ferry Building,
+Suite 200, San Francisco, 94111.
+    "Software" shall mean the MonoDepth v2 Software uploaded by Licensor to the
+GitHub repository at https://github.com/nianticlabs/monodepth2
+on May 2nd 2019 in source code or object code form and any
+accompanying documentation as well as any modifications or additions uploaded
+to the same GitHub repository by Licensor.
+
+
+2.  License.
+
+    2.1 The Licensor has all necessary rights to grant a license under: (i)
+copyright and rights in the nature of copyright subsisting in the Software; and
+(ii) certain patent rights resulting from a patent application filed by the
+Licensor in the United States in connection with the Software. The Licensor
+grants the Licensee for the duration of this Agreement, a free of charge,
+non-sublicenseable, non-exclusive, non-transferable copyright and patent
+license (in consequence of said patent application) to use the Software for
+non-commercial purpose only, including teaching and research at educational
+institutions and research at not-for-profit research institutions in accordance
+with the provisions of this Agreement. Non-commercial use expressly excludes
+any profit-making or commercial activities, including without limitation sale,
+license, manufacture or development of commercial products, use in
+commercially-sponsored research, use at a laboratory or other facility owned or
+controlled (whether in whole or in part) by a commercial entity, provision of
+consulting service, use for or on behalf of any commercial entity, and use in
+research where a commercial party obtains rights to research results or any
+other benefit. Any use of the Software for any purpose other than
+non-commercial research shall automatically terminate this License.
+
+
+      2.2 The Licensee is permitted to make modifications to the Software
+provided that any distribution of such modifications is in accordance with
+Clause 3.
+
+      2.3 Except as expressly permitted by this Agreement and save to the
+extent and in the circumstances expressly required to be permitted by law, the
+Licensee is not permitted to rent, lease, sell, offer to sell, or loan the
+Software or its associated documentation.
+
+
+3.  Redistribution and modifications
+
+    3.1 The Licensee may reproduce and distribute copies of the Software, with
+or without modifications, in source format only and only to this same GitHub
+repository , and provided that any and every distribution is accompanied by an
+unmodified copy of this License and that the following copyright notice is
+always displayed in an obvious manner: Copyright © Niantic, Inc. 2018. All
+rights reserved.
+
+
+    3.2 In the case where the Software has been modified, any distribution must
+include prominent notices indicating which files have been changed.
+
+    3.3 The Licensee shall cause any work that it distributes or publishes,
+that in whole or in part contains or is derived from the Software or any part
+thereof (“Work based on the Software”), to be licensed as a whole at no charge
+to all third parties entitled to a license to the Software under the terms of
+this License and on the same terms provided in this License.
+
+
+4.  Duration.
+
+    This Agreement is effective until the Licensee terminates it by destroying
+the Software, any Work based on the Software, and its documentation together
+with all copies. It will also terminate automatically if the Licensee fails to
+abide by its terms. Upon automatic termination the Licensee agrees to destroy
+all copies of the Software, Work based on the Software, and its documentation.
+
+
+5.  Disclaimer of Warranties.
+
+    The Software is provided as is. To the maximum extent permitted by law,
+Licensor provides no warranties or conditions of any kind, either express or
+implied, including without limitation, any warranties or condition of title,
+non-infringement or fitness for a particular purpose.
+
+
+6.  LIMITATION OF LIABILITY.
+
+    IN NO EVENT SHALL THE LICENSOR AND/OR AUTHORS BE LIABLE FOR ANY DIRECT,
+INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING
+BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
+ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+7.  Indemnity.
+
+    The Licensee shall indemnify the Licensor and/or Authors against all third
+party claims that may be asserted against or suffered by the Licensor and/or
+Authors and which relate to use of the Software by the Licensee.
+
+
+8.  Intellectual Property.
+
+    8.1 As between the Licensee and Licensor, copyright and all other
+intellectual property rights subsisting in or in connection with the Software
+and supporting information shall remain at all times the property of the
+Licensor. The Licensee shall acquire no rights in any such material except as
+expressly provided in this Agreement.
+
+    8.2 No permission is granted to use the trademarks or product names of the
+Licensor except as required for reasonable and customary use in describing the
+origin of the Software and for the purposes of abiding by the terms of Clause
+3.1.
+
+    8.3 The Licensee shall promptly notify the Licensor of any improvement or
+new use of the Software (“Improvements”) in sufficient detail for Licensor to
+evaluate the Improvements. The Licensee hereby grants the Licensor and its
+affiliates a non-exclusive, fully paid-up, royalty-free, irrevocable and
+perpetual license to all Improvements for non-commercial academic research and
+teaching purposes upon creation of such improvements.
+
+    8.4 The Licensee grants an exclusive first option to the Licensor to be
+exercised by the Licensor within three (3) years of the date of notification of
+an Improvement under Clause 8.3 to use any the Improvement for commercial
+purposes on terms to be negotiated and agreed by Licensee and Licensor in good
+faith within a period of six (6) months from the date of exercise of the said
+option (including without limitation any royalty share in net income from such
+commercialization payable to the Licensee, as the case may be).
+
+
+9.  Acknowledgements.
+
+    The Licensee shall acknowledge the Authors and use of the Software in the
+publication of any work that uses, or results that are achieved through, the
+use of the Software. The following citation shall be included in the
+acknowledgement: “Digging Into Self-Supervised Monocular Depth Estimation,
+by C. Godard, O. Mac Aodha, M. Firman, G. Brostow, arXiv:1806.01260”.
+
+
+10. Governing Law.
+
+    This Agreement shall be governed by, construed and interpreted in
+accordance with English law and the parties submit to the exclusive
+jurisdiction of the English courts.
+
+
+11. Termination.
+
+    Upon termination of this Agreement, the licenses granted hereunder will
+terminate and Sections 5, 6, 7, 8, 9, 10 and 11 shall survive any termination
+of this Agreement.