Skip to content

Palm Vein Identification using CASIA Database, C# and TensorFlow

License

Notifications You must be signed in to change notification settings

Seregamil/Palm-Vein-Identification

Repository files navigation

Palm Vein Identification

Development is in C# with .net interactive Database CASIA Multi-Spectral Palmprint Database: http://biometrics.idealtest.org/dbDetailForUser.do?id=6

ROI extracting method: jupyter notebook (.net interactive) On average, it took about 2 minutes to process 7200 images.

Due to the fact that the data set is rather poor for training, an attempt was made to artificially increase the amount of data. To do this, the original image is rotated by a certain angle and the ROI is extracted, expanding the data set.

Source dataset: 6 images on one palm on one frequency.
Dataset now: 1080 images on one palm on one frequency.

Image rotation angles: 0, 2, 4, ... 358
Image processing time (1 hand - 940nm): 04:51:25.1516273

The total number of output images: 108000

Model accuracy for 108000 images: ??.??%
Elapsed training time for 108000 images: ??d ??h ??m

Model accuracy for 12600 images: 64,20%
Elapsed training time for 12600 images: 11h 20m

Packages

  1. OpenCvSharp4.Windows
  2. ML.Net
  3. TensorFlow

About

Palm Vein Identification using CASIA Database, C# and TensorFlow

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published