In the research paper, Histograms of Oriented Gradients for Human Detection, the images used are of 64X128 pixels. Do we need to crop our images dataset for applying histogram of oriented gradient?
I agree with Vadym. I have used the same size of the dataset with FFT.
The choice of 64X128 pixels images has some aspects.
As the first aspect need to mark that such size is applicable to use Fast Furie Transform (FFT) for spectral preprocessing and use fast convolution CNN.
On the other hand, this size is a result of the limitation of calculation performance, because authors of this paper wrote:
"Our 64×128 detection window includes about 16 pixels of margin around the person on all four sides. Fig. 4(e) shows that this border provides a significant amount of context that helps detection. Decreasing it from 16 to 8 pixels (48×112 detection window) decreases performance by 6% at 10−4 FPPW. Keeping a 64×128 window but increasing the person size within it (again decreasing the border) causes a similar loss of performance, even though the resolution of the person is actually increased ".
I agree with Vadym. I have used the same size of the dataset with FFT.
I have using dlib library for detection face, in the program is using 64x128 pixel for computing,
maybe you will find the answer from Dalal thesis, so I think it will easy to compute with sobel operator
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing