Simple HTTP Image Server

A drop-in replacement for python -m http.server, albeit for images.

Quickstart

Install shis.

$ pip install shis

Navigate to a directory/containing/images.

$ cd /directory/containing/images

Remember python -m http.server? Great.

$ python -m shis.server
# Serving HTTP on 0.0.0.0:7447. Press CTRL-C to quit.
# Processing images from : directory/containing/images
# Creating thumbnails in : directory/containing/images/shis
# Generating Website     : 100%|████████████████████|     4/    4 [00:00<00:00,  49.82it/s]
# Generating Thumbnails  : 100%|████████████████████|   300/  300 [00:00<00:00,  1.47kit/s]

There. You can now head over to http://0.0.0.0:7447/ (Or use your public IP instead).

TIP: You can install the latest development version directly from GitHub.

$ pip install git+https://github.com/nikhilweee/shis/

Preview

Here's an example of what you can expect to see. A live preview is also available at nikhilweee.github.io/shis.

Features

• Drop-in replacement for python -m http.server, so it's easy on your brain.
• Serves website even before creating thumbnails, so you don't have to wait.
• Uses multiple processes to create thumbails, so it's fast.
• Efficient resumes, so we build on past progress.
• Creates both small and large size thumbnails, so it's easy on your eyes.
• Minimal dependencies - just Pillow, Jinja2 and tqdm.
• Server side pagination, so it's easy on your browser.
• Tries to preserve EXIF orientation, so you don't have to rotate manually.
• Displays the public IP (if exists), so you don't have to remember.

Usage

The following options are available. You can also access this via python -m shis.server -h. Further documentation can be found at shis.readthedocs.io.

usage: python -m shis.server [-h] [--image-dir DIR] [--thumb-dir DIR]
                             [--port PORT] [--pagination ITEMS] [--order ORDER]
                             [--ncpus CPUS] [--clean] [--previews]
                             [--thumb-size SIZE] [--preview-size SIZE]

A drop in replacement for python -m http.server, albeit for images.

optional arguments:
  -h, --help            show this help message and exit
  --image-dir DIR, -d DIR
                        directory to scan for images (default: current dir)
  --thumb-dir DIR, -s DIR
                        directory to store generated content (default: shis)
  --port PORT, -p PORT  port to host the server on (default: 7447)
  --pagination ITEMS, -n ITEMS
                        number of items to display per page (default: 200)
  --order ORDER, -o ORDER
                        image listing order: name (default), random, original
  --ncpus CPUS, -j CPUS
                        number of workers to spawn (default: available CPUs)
  --clean, -c           remove existing --thumb-dir (if any) before processing
  --previews, -f        also generate fullscreen previews (takes more time)
  --thumb-size SIZE     size of generated thumbnails in pixels (default: 256)
  --preview-size SIZE   size of generated previews in pixels (default: 1024)

Benchmarks

For comparison, I ran the following tools on the FFHQ Dataset. The dataset contains 70k images in 1024x1024 resolution for a total size of 90GB. The converted thumbnail size was set to 320x320 for all tools. The tests were done on a machine with an AMD EPYC 7401P CPU with 24 Cores, 32GB Memory and Python 3.6.10 running on Ubuntu 18.04. The config files used are provided below. All conversion times are in hh:mm:ss format.

Why SHIS?

There are a bunch of static image servers (thumbsup, sigal, etc) available in a bunch of different languages (javascript, python, etc). While some of them like fgallery and curator haven't been developed in a while, others like thumbsup and sigal take a lot of time converting images. SHIS is designed with just one use case in mind, and it plans to do it well. It aims to serve a large directory of images in the fastest and easiest way possible.

Acknowledgements

The demo at nikhilweee.github.io/shis shows sample images from the ImageNet dataset obtained through https://github.com/EliSchwartz/imagenet-sample-images. The gallery template used to display images is a modified version of the cards theme from https://github.com/thumbsup/theme-cards.

License

MIT License