I have currently a test case with a 140 000 images index. The search time is less than 2 seconds on 3.00GHz Core2 Duo CPU with 4 GB of RAM. The accuracy is then, for me, satisfying . Pastec is still being developed and there are many possible improvements to reduce this time.
It is however difficult to compare Pastec with the competition as most image recognition APIs are not free for large databases.
Pastec only allows to recognize images. You provide reference images with their corresponding unique integer identifiers. Pastec extracts features for each reference image and stores them in its database.
When you perform a request, Pastec compare the features of the request image with the ones stored in the index. If it detects that the request image is similar enough to one or more images of the index, then it returns their identifiers.
You can for example use Pastec to recognize DVD, CD or Book covers in a mobile app.
Currently, it does not allow to extract text from images.
The documentation indeed needs improvements. I will work on it.
In a nutshell at this moment it is classification tool, but not based on pre-trained model. So, when it looks for feature index does it pick the first available (or) it iterates for best available? good luck!
Inside Pastec, the images are represented using the visual word paradigm: http://en.wikipedia.org/wiki/Visual_Word
During indexing, each image feature is assigned to the nearest visual word among a pre-trained set of 1 million.
About 1000 visual words are extracted per image.
Thank you for your comment.
I have currently a test case with a 140 000 images index. The search time is less than 2 seconds on 3.00GHz Core2 Duo CPU with 4 GB of RAM. The accuracy is then, for me, satisfying . Pastec is still being developed and there are many possible improvements to reduce this time.
It is however difficult to compare Pastec with the competition as most image recognition APIs are not free for large databases.