This project is posted for the user "mastermatlab ", following our discussion on this project: [url removed, login to view]
**** Please do not bid on this project, as the contractor has already been chosen ****
The project is to scan many photos and to define which one includes people in swimsuits. Those who are identified as containing swimsuits will be moved to a different folder (called "swimsuit" or "positive").
There are many regular photos (without swimsuits) and few swimsuit photos, so it is quite challenging to discover [url removed, login to view] ratio is about 15:1 or 20:1.
There are 2 measures to test the results:
1) the success ratio: out of 100 swimsuit photos, how many were correctly discovered as such.
2) the efficiency: out of all the photos that were discovered as swimsuits, how many are actually swimsuits? for example, lets say we discovered 100, and only 50 are actual swimsuits (50 mistakes). the efficiency ratio is 50%.
Project key points:
1. required efficiency ratio: 90% or more. required success ratio: 60% or more. ratios defined above.
2. Those ratios will be measured on a set of photos that I will provide you. The set will include ~17,000 normal photos and ~1,000 swimsuit photos.
3. The code will have to scan all the files in a specific folder (the folder will be constantly updated), analyze the photos, and then move the photo to the correct folder: if this is a swimsuit photo, move it to the "swimsuit" folder, otherwise move it to the "regular" folder, or delete it.
4. The code will be installed on a UNIX server, and fully operational.
Two milestones: (1) full operational software with the desired ratios (90% efficiency and 60% success), measured on the sample;and (2) installation on the UNIX server.
Agreement with Mastermatlab: the installation on the UNIX server (part 4) is NOT a part of the project, and will be done in a seperate project. There are no milestones for this project, it is a one-deliverable project.
The deliverable is a matlab code that runs on a folder with unlimited number of photos (the test sample will be ~18,000), scans them and moves the photos to correct folders, as described above. The code / algorithm must meet the target ratios.