Skip to content

Latest commit

 

History

History
59 lines (46 loc) · 1.82 KB

README.md

File metadata and controls

59 lines (46 loc) · 1.82 KB

dejavu -- distributed image recognition

dejavu uses the TeleHash DHT to ditribute pictures and match them against each other.

One use case would be for a search and rescue team to identify victims in a post-disaster situation. Given each member of the team has a part of the database of faces such that the entire network contains the whole database, a face can be effeciently recognized by distributing a picture among the peers.

daemon

The TeleHash and HTTP daemon which broadcasts and listens to app-specific communication.

Use TeleHash to broadcast signals and commands, HTTP to serve the images, frontend and API.

telehash

photo telex -- always contains url(http), and hash of the image, optionally contains a tag (an object identifying the photo)

signals and commands that can be attached a photo telex to:

  • .store -- store a reference image, an original
  • +match -- a photo of a person to be ID'ed
  • .ack -- acknowledgement for staging of a photo for recognition or after storing it retry if not received within a timeout

when the daemon is started with the --seed option, it skips bootstrap. It's world-facing IP has to be manually determined and distributed to the new coming peers.

web interface

Easy web based interface that allows for drag-drop input, shows percentage similarity, allows humans to attest results. Preferably implemented in pure node.js.

api

Simple RESTful HTTP API for mobile apps to use.

files

bin/
  `--dejavu        # executable to start the daemons
lib/
  `--ext/          # external libraries, in 'paths'
  `--dejavu/       # our libraries
  `--dejavu.js     # setup
test/              # tests