All projects17 / 43
AI & Data
Leaf Recognition
Shape & texture computer vision
At a glance
- Discipline
- AI & Data
- Key features
- 5 delivered
- Screens
- 4 in gallery
PythonOpenCVNumPyFourier TransformLBP
Leaf Recognition
Overview
A classical vision pipeline using a Multiscale Triangle Descriptor and LBP features, reaching 87.61% top-1 accuracy across 100 species.
Images are denoised, thresholded with Otsu and refined morphologically before contour extraction. Fourier analysis turns shape signals into rotation-invariant descriptors.
Texture is captured with uniform Local Binary Pattern histograms, and both feature sets are combined for 1-NN classification with no high-level ML libraries.
Key features
- 01Custom Multiscale Triangle Descriptor
- 02Rotation-invariant Fourier features
- 03LBP texture histograms
- 0487.61% accuracy on CVIP100
- 05Five bootstrap validation runs
Gallery
Inside the product
4 screens
Up next18 / 43
PPE Safety Detection
Real-time compliance monitoring
AI & DataPPE Safety Detection