photolib/README.md

Photolib
========

Cloud-native photo and related media library software, with a focus for photography hobbyists.

Photolib provides a simple, rich, and modern UI to manage your photo and video library. It can scale to host hundreds of
thousands of photos but is also light and simple enough to host just the couple photos you shoot with your smartphone.

Photolib stores images as a logical "photoset". This allows identical images and metadata to be grouped into one item.
Many photographers shoot in multiple formats - that is, to produce multiple files per click of the shutter. A JPEG and
a RAW of some sort (Like Canon's CR2) is common. The primary idea behind Photolib's organization is that one click of a
camera shutter equals one photoset.

When photos are imported - assuming they are done in the same batch - files are grouped based on their name and
extension, using rules that apply to almost every camera format. For example, a camera shooting under the "multiple
formats" mode described above would create, from one shot, `IMG_1234.JPG` and `IMG_1234.CR2`. Photolib would detect that
since these two images have the same file name (the `IMG_1234` part), they are the same photo and should be grouped into
a photoset.

In addition to controls in the web interface, Photolib provides both a REST API and command-line tools to manage your
photo library.


Installation
------------

Pre-built releases are not currently available. See `Dockerfile` for a step-by-step installation from source.

The general steps are:

* `npm install` - install modules needed to build the frontend assets
* `grunt` - build the frontend assets
* `pip3 install -r requirements.txt` - install python dependencies
* `python3 setup.py install` - install Photolib and command line tools

Note that the npm and grunt steps can be skipped if you're not planning to run the web UI (e.g. you want only the
command-line tools).


Usage
-----

After installation, run the server:

* `photoappd --port 8080 --library file://./library --database sqlite:///photos.db --cache ./cache`

Arguments are as follows:

* `--library file://./library` - file storage uri, in this case the relative path `./library`
* `--database sqlite:///photos.db` - [Sqlalchemy](https://docs.sqlalchemy.org/en/13/core/engines.html) connection uri
* `--cache file://./cache` - storage uri to use as a cache for things like thumbnails. It can be the same as the library
* `--port 8080` - listen on http on port 8080
* `--thumb-service` - optional address of thumbnail service - see below.

Supported library uri schemes are:

* file - relative (as above) or absolute file paths - `file:///srv/library`.
* minio - `minio://username:password@host/bucket_name/path/prefix`

Photolib is tested with [Minio](https://min.io/), but should work with S3. For the database, minio is tested with SQLite
and MySQL (using the `mysql+pymysql://` driver).

Next, the `photousers` command can be used to create a user account. Login information is necessary to see images marked
as private or upload images. You may want to run this before starting the server, but either order works.

* `photousers create -u dave -p mypassword`

That's it - you've set up Photolib! Next steps from here would be to the other command line tools to import media.

Besides browsing, most interaction with Photolib will be done using its CLI tool, `photocli`. Note that server address,
username, and password details need to be provided to the cli via a url flag not shown here. See `photocli --help` for
more information.

Optional: Photolib uses a secondary service to generate thumbnails for video files. After the above installation
instructions are complete, run `photothumbd` on a differnet port with the same arguments. On first run, it will create
a user for internal communications and log the username/password or you can set `THUMBSERVICE_INITIAL_PASSWORD_HASH` to
the sha256 hash of the desired password. This username/password must be used in the `--thumb-service` option
for `photoappd`.


Commands
--------

* `photocli ingest` - import new photos into the library:


```
$ photocli ingest photos/*.jpg raws/*.cr2
```

or add a new format to an existing photo:

```
$ photocli ingest -c 52679a30-cee4-45c4-8d83-80ee92242e1c photos/*.jpg
```


* `photocli checkdupes` - scan files and print those not already in the library:


```
$ find source -type f -exec shasum -a 256 {} \; | tee shas.txt
544b5a4f3898abae73260930ee70931a3992c048649692cef89b37576e886f69  source/2018-12-09 13.52.22.jpg
d2c2c30ca4986d364026644881d667162031008e60aac6a69d7b18230b7ea98c  source/2019-04-08 20.05.02.jpg
9d42859ed92d7fb978cf73b41293480e1eab03d2d3a14c185c4daf3a49d324ab  source/2019-06-19 16.28.07.png
...

$ photocli checkdupes --sha-files shas.txt
```

TODO: support not using `--sha-files`.

This command, along with `ingest`, allows for fun tool combinations like so:

```
photocli checkdupes --sha-files shas.txt | xargs -d "\n" photocli -y ingest
```

This would ingest all the files listed in `shas.txt` that aren't already in the database.


Roadmap
-------

- TODOs:
    - need `nullable=False` on columns where applicable
    - 0-5 rating option for photosets
    - give uuids a style like `this`
    - search next/prev buttons to match feed links
    - search tag selector to match tag picker links
    - <select> styling
    - improve stats page to differentiate PhotoSets and image files
    - improve stats page to have yearly totals or even a daily view on a calendar
    - redo feed pager into a <form> so things like `pgsize` stick
    - portrait/landscape filter in search
    - DARK MODE
    - favicon
    - indicator in thumbnails if something is a video, particularly in search
    - search by type of image or video
    - 'preview' link on photo view web pages seem to 500
    - some video files (that ffmpeg can generate a thumb from) get mime'd as 'audio/mp4'
    - videos under 1s fail to generate a thumb (because of the hard-coded 1s thumb extraction point)
    - link on the photo's minimap that takes you to the main map at the location of that photo
    - nullable=False for all columns possible
    - check/clean no-js fallbacks
    - check gps skew in "nearby" section
    - camera info / database
        - show the camera information on each photo page
        - create a catalog of known cameras by extracting the info from the exif
        - associate cameras with photos and allow search by camera
        - cameras can be nicknamed
        - all of this with lenses also?
    - make it clearer that the download/preview button gets a (converted) original size image
    - make the map view markers colored in some logical manner
        - rainbow across age spectrum?
        - 4 colors (with shades) for the 4 seasons?
        - colorblindness?

- Flesh out CLI:
    - data manipulation
        - combine photosets
        - split photosets
    - Relink function - make a photo a member of another photo
    - Config that is saved somewhere
    - Album features
    - Support additional fields on upload like title description tags etc
    - delete features
    - tag features
        - Tag on import
    - modify features (tags & images)
    - "Batch" tag on import
        - Generate a tag on import
        - Save it in config and re-use it (if passing --same-batch)
        - photos imported as a batch will be under 1 tag

- Longer term ideas:
    - "fast ingest" method that touches the db/storage directly. This would scale better than the API ingest.
    - Dynamic svg placeholder for images we can't open
    - Proactive thumbnail generation
    - on photo thumbs on the feed, a little "2" indicating there are 2 images in the set (or w/e number for that item)
    - dark theme
    - more information from the images like on http://exif.regex.info/exif.cgi
    - Map page: use the craigslist style map groups?

- my photos problems
    - recently imported imotrip photos not time offset lol
        - time offset options in ui?

- Low-priority TODOs
    - Cli functions to download images
    - feature to "download a web safe/optimized copy" - get a compressed jpeg at a common res like 1024x768 with exif stripped


Facial Recognition
^^^^^^^^^^^^^^^^^^

Proposed feature. Not yet under development. This section contains notes about its development.

Prior work / modules to use:

* https://github.com/ageitgey/face_recognition/blob/master/examples/find_faces_in_picture.py
* https://face-recognition.readthedocs.io/en/latest/face_recognition.html
* https://www.pyimagesearch.com/2018/06/18/face-recognition-with-opencv-python-and-deep-learning/
* https://github.com/JohannesBuchner/imagehash


Image processing framework
^^^^^^^^^^^^^^^^^^^^^^^^^^

Purpose:

Some way to mass process photos in batches. For example, the batch jobs could:
    - load metadata from images that is newly supported
    - process new faces in images
    - create a downloadable mini-album
    - proactively generate thumbnails
    - replace thumbservice, the video thumb service
    - import photos
        - also by grabbing them from a foreign source e.g. sftp

Could also have:
    - a session-based mechanism to "select" photos
        - e.g. to add them to a batch processing job or album later. Like a clipboard of sorts

Design notes:
    - general ui:
        - use a tag multiselector (like the search page) to pick the "input" images for a batch job
        - for now, we can have a drop-down to select which job to run
        - there could then be a second-stage config screen for the job where
            - additional parameters for the job could be set
                - like what?
            - the job could run some sort of validation
        - there is a Job List screen that displays an index of current / past jobs

    - job executor:
        - "for now", a SimpleExecutor service.
        - it's an http service (like thumbservice lol) that is similarly notified by http hook when a job is created by
          the ui/api (except we pass the ID in the http call this time and it looks up the details from the DB)

    - internal notes:
        - tables / cols:
            - Job: id, uuid, name?, job_name, status(new, running, done, error)
            - JobTarget: id, job_id, type(tag, set, photo), target
                - this indicates where a job should get images to work on
                - there is a numeric ID in the `target` column that can be any of a Photo, PhotoSet, or Tag
                - which it is, is determined by the `type` column
                - therefore the code can know which method to locate the file with

    - need testing to ensure the program is usable without having a dedicated executor etc