I am curious as to how I can access the database the scrapers use (and build) on my OSMC client. I can shell into it, a Raspberry Pi2, but attempts to access any databases on that unit via MySQL Workbench have failed.
I want to copy off that database and use it to add features, either via an addon, or otherwise, that give me access to additional info from a playing movie… like actors staring in that movie, what they were paid to be in it, what their net worth is, who their agent is, what new work are they going to do, etc. Whatever I can come up with that my family grabs their tablets to research on.
I have no clue how to write an addon, yet, but want to add these informational features to the GUI if possible, and maybe even try facial recognition from a still frame to identify an unknown actor. Yeah, that last one is pretty far out there as of yet.
Reason I went this route, with OSMC, Linux, Pi2, and other Linux Embedded devices I have running all over my house, in the walls, in the roof, in my Jeep. I want control of the content my system presents.
Pertaining to such specifics regarding internal functions of Kodi, you’re more likely to get better and faster info by seeking such info from Kodi themselves. Most OSMC users are just that, users. Your questions pertaining specifically to Kodi will get a lot more visibility by Kodi dev’s on their forums. Just a tip…
Thank you guys, that is what I was looking for. Already have a MySQL server running on two of my home servers. One is more media related, and the other is more data (budget, data acquisition, security cameras, home automation). It’s a pair of busy database servers as they handle a ton of logging and query traffic from my embedded systems, sensor nodes, cameras, whatever. The list is pretty long. I have never had time to go after scraping, and what OSMC has done with my media library is outstanding! I just want control and mod ability with that as I do with ALL things I acquire data about. Anyone that does what I do would.
Thanks for the link breadcrumb trail. I shall follow.
You could do this quite easily. Either invoke Kodi’s screenshot function (which uses Dispmanx), or use Dispmanx directly. Crop the photo accordingly to only show faces using OpenCV and then fire that off to Google’s reverse image search.
Yeah, I figured Google would be the option. I would love to have a local option for that sort of work though. Take the auto-cropped image and compare it with a proprietary database of select images, like those populated by scraping, and find a high-probability match. The possibilities controlled by me as to the size of the image database. The truth is out there, I just need to find it.
I find that the gift I have of identifying an actor centers on my ability to recognize the similarities with distance apart of the eyes, shape of the eye socket, eyebrows, nose, length of the face, etc. If I can duplicate that algorithm that my mind uses to astound my family, in code, write the script to look for what I do unintentionally, then I am set. Sorry for the off-topic extra content.
I found a few promising open source libraries for face recognition that can be built in Debian. I am going to try these against still frames, using a small list of actor images. The idea is to send the image to the home server, an I7 Quad Core, have the image analyzed and process it using an array of algorithms against a database of portraits, then have the probabilities crunched and a result returned. This result will be processed at the client level to identify the actor with a probability score.
It sounds like black ops stuff, at least to me, but would be a cool feature to use while watching a movie, and having the results validated against a movie database, which is why I asked about database access in the first place.
