- Sequentially version by numba no python mode.
- Parallel version by numba cuda mode.
- Optimize using shared memory and experiment with different block sizes for kernel functions
- seam_carving_cpu.py [-dx <seam_dx>] [-dy <seam_dy>] [-in <image_in>] [-out <image_out>] -test_time
- dx: Number of horizontal seams to add (if positive) or subtract (if negative)
- dy: Number of vertical seams to add (if positive) or subtract (if negative)
- in: input image file path
- out: output image file path
- test_time: print time of each step in pipeline
- seam_carving_gpu.py [-dx <seam_dx>] [-dy <seam_dy>] [-in <image_in>] [-out <image_out>] -check_sum
- dx: Number of horizontal seams to add (if positive) or subtract (if negative)
- dy: Number of vertical seams to add (if positive) or subtract (if negative)
- in: input image file path
- out: output image file path
- check_sum: use cpu's grayscale