README.md

In mdlincoln/PastecR: Wrapper for the Pastec Image Search and Recognition Engine API

PastecR

Wrapper for the Pastec image search and recognition engine API.

Directions for building and running Pastec can be found here. (Note that if you are trying to install Pastec on OS X or Ubuntu, you may find that this fork does better at locating the necessary libraries.)

Once you have an instance of Pastec running, you can start interacting with it using PastecR.

Installation

# install.packages("devtools")
devtools::install_github("mdlincoln/PastecR")
library(PastecR)

Usage

Pastec is an open-source image matching engine that does Google-images-esque fuzzy searching of a database of images that you compile. It is very good at finding close-but-not-exact matches, such as different versions of a photograph, or rotated, cropped, or resized versions of images.

PastecR provides a wrapper around the Pastec API, which can be particularly useful when parsing the results returned by an image search request.

Self-hosted or Saas?

You can compile and run Pastec yourself, however there is also a hosted version available with both a free and a paid option at https://api.pastec.io. PastecR supports both options:

library(PastecR)

# open_pastec_server defaults to the same port defaults as Pastec
ops <- open_pastec_server(url = "localhost", port = 4212)

# You can find the index_id and auth_key for your api.pastec.io account at
# <https://api.pastec.io/manager/apiPage>
hps <- hosted_pastec_server(index_id = "XXXXXXXXXXXXXXXXXXXX", auth_key = "XXXXXXXXXXXXXXXXXXXX")

Load images and search

We'll be using a number of artowrks for this example, including three different versions of a print.

Once you have established either your own instance of Pastec, or an account at https://api.pastec.io, you may add images with the add_image() command. You must set an integer image_id when adding.

# Get image paths
erasmus1 <- system.file("img", "RP-P-1906-1485.jpg", package = "PastecR")
erasmus2 <- system.file("img", "RP-P-1906-1486.jpg", package = "PastecR")
erasmus3 <- system.file("img", "RP-P-1906-1487.jpg", package = "PastecR")
nightwatch <- system.file("img", "SK-C-5.jpg", package = "PastecR")

add_image(erasmus1, 1, ops)
add_image(erasmus2, 2, ops)
add_image(erasmus3, 3, ops)
add_image(nightwatch, 4, ops)

# Images can also be removed by index
remove_image(4, ops)

Once your index is loaded, you can search for similar images, and parse the results with a number of helper commands.

search_result <- search_image(erasmus1, ops)
results_as_data_frame(search_result)

##   height width  x   y image_id scores tags
## 1    827   592 53 142        1    964   NA
## 2    803   567 64 159        3     62   NA
## 3    796   575 56 173        2     59   NA

See ?pastec_response for other parsing functions.

Index operations

To clear the index, use clear_index(server). On the open, self-hosted version of Pastec, you can also save and load index files to disk by using save_index() and load_index(). At the moment, these commands are not available for the SaaS version of pasted, and will throw an error.

Matthew Lincoln

mdlincoln/PastecR documentation built on May 22, 2019, 4:03 p.m.