Frank Reilly Method Fonte: PinterestDescrição completa
Frank Reilly Method Fonte: PinterestFull description
An Atlas of the human anatomy, for artists.
Description complète
An Atlas of the human anatomy, for artists.Descripción completa
Descrição completa
Full description
An Atlas of the human anatomy, for artists.Description complète
An Atlas of the human anatomy, for artists.Full description
An Atlas of the human anatomy, for artists.
Full description
Full description
Guitar MethodFull description
Descrição completa
Pianist Barry Harris' method which outlines a more summarized, yet in-depth, approach to jazz harmony. Method adapted for guitar by Alan Kingstone.
Survey of Human detection Student Name: Nguyen The Manh Supervisor: Dr. Vo Le Cuong
Content
Previous week: Survey some aspects related to human detection Tools and data sets for human detection and evaluation methods Overview of HOG method Work of current week Study detail implementation of HOG method Refer to some HOG implementations in MatLab Learn basic OpenCV for image processing
As the authors of [2] point out, these normalizations have only a modest effect on performance, because the subsequent descriptor normalization essentially achieves the same result.
2. Gradient Computation
The HOG uses 64x128 detection window The gradient is computed for each pixel in 8x8 pixel cells The most common method is to apply the 1-D centered, point discrete derivative mask in one or both of the horizontal and vertical directions: [-1, 0, 1] and [-1, 0, 1] T
64x128 pixel detection window 8x8 pixel cell
3. Orientation Binning
64 gradient vectors in each 8x8 pixel cells is put them into a 9-bin histogram ranges from 0 to 180 degrees. Each pixel calculates a weighted vote for an edge orientation histogram channel based on the orientation of the gradient element centered on it. The author [2] found that unsigned gradients with 9 histogram channel performed best in their experiment.
4. Descriptor Block
Grouping cells into larger spatial block and normalizing each block separately. The block usually overlap so each cell response contributes more than once to the final descriptor.
The final HOG descriptor is the concatenated vector of the components of the normalized cell histograms from all of the block regions. The author [2] evaluated 2 classes of block geometries: R-HOG: rectangular HOG C-HOG: circular HOG
5. Block normalization
Concatenating the histograms of the 4 cells within each block into a vector with 36 components (4 histograms x 9 bins per histogram) and normalize each block separately. The author [2] evaluated four different block normalization schemes: L2-norm L2-Hys L1-norm L1-sqrt
SVM Classifier
The author [2] tested 2 SVM classifiers: Soft linear SVM Gaussian kernel SVM: increase performance but higher run time
Current problems
Have not understand clearly the implementation of computing and normalizing the block description. Not have background knowledge of SVM to understand the classifier process. Have not understand the evaluation method of the HOG method in [2] Due to above problems, I have not implement HOG method successfully.
Future Work Work for next week: Continue studying the HOG method to clarify unclear points Implement HOG method in MatLab.
References
[1] Nguyen, Duc Thanh, Wanqing Li, and Philip O. Ogunbona. "Human detection from images and videos: A survey." Pattern Recognition (2015). [2] Dalal, Navneet, and Bill Triggs. "Histograms of oriented gradients for human detection." Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on. Vol. 1. IEEE, 2005. [3] https://chrisjmccormick.wordpress.com/2013/05/09/hog-persondetector-tutorial/ [4] https://en.wikipedia.org/wiki/Histogram_of_oriented_gradients
