digiKam 7.0.0-beta1 released, digiKam is a digital photo management tool for the KDE desktop environment, this version brings a few highlights.
After the face of deep learning management functions
Over the years, digiKam provides an important function, dedicated to the detection and face recognition photo. Since the first revision (digiKam 2.0.0) contains this feature, the background algorithm used (not deep learning algorithm based) older and not change. Its function is not sufficient to automatically promote human face management workflow.
Up to now, using the OpenCV library based on the classic characteristics of Cascade classifier performing a complex method for analyzing the image content to isolate and marked face. This older method is feasible, but does not provide a high level of true positive results. In the analysis, face detection capable of providing 80% of the good results, although not bad, but it requires a lot of user feedback to determine whether a face is detected. And, in the user feedback, face recognition function can not provide a good experience for automatic tagging mechanism for the user.
This year, the function of the code is completely rewritten. The project's goal is to keep all the old ideas, and detection and recognition engine ported to more modern deep learning method. Based on the latest OpenCV library in the new code of neural network function of the depth of the neural network data model learned in advance is dedicated to face management. You do not need the learning phase can be carried out face detection and recognition. The result is faster, the success rate as high as 97%. In addition, the non-face can also be detected, like a dog in this screenshot to see the same.
In the future, the new code structure will allow the use of other pre-learned data model to detect and identify the image in a form (e.g., monuments, plants, animals). The difficulty is to find a suitable pre-learning neural network data.
The new RAW file support
digiKam tries to bring all file functions provided by the most powerful digital camera. Raw file support is a huge challenge. Therefore digiKam use powerful Libraw library after Raw files on the computer for processing, the library contains sophisticated algorithms to support all the different Raw file formats.
This 7.0.0 version with the new libraw 0.20, which was introduced more than 40 new Raw format, is available on the market, especially the latest camera models, including the new Canon Canon CR3 format and Sony A7R4 (61 Mpx!).
7.0.0 official version is expected to be released in the spring of next year.
Update Description: https://www.digikam.org/news/2019-12-22-7.0.0-beta1_release_announcement/