| 128 | |
| 129 | == 4 June -- 11 June week 3: implement mean-shift image segmentation algorithm == |
| 130 | === What did you get done this week? === |
| 131 | *Implement the mean-shift algorithm iteration part [https://trac.osgeo.org/grass/browser/sandbox/bo/i.segment.gsoc2016/i.segment/mean_shift.c mean-shift]. |
| 132 | |
| 133 | === What do you plan on doing next week? === |
| 134 | * Previously there are 3 tasks planning for the past week, but because of the unexpected debugging time only one task finished for this week, I will try to finish the following two tasks in the next week. |
| 135 | |
| 136 | 1. Refer to the r.clump/clump.c to implement the object segmentation part. |
| 137 | |
| 138 | 2. Merger the smaller super-pixels to adjacent super-pixel which under the threshold and write to the output |
| 139 | |
| 140 | === Are you blocked on anything? === |
| 141 | * Some parameters could not pass from parse_args.c to mean_shift.c, for example: |
| 142 | {{{ |
| 143 | globals->ms_spatial_bandwidth and globals-> ms_range_bandwidth |
| 144 | }}} |
| 145 | |
| 146 | * There is a random appeared bug, sometimes it can output normally to console, but some time the output doesn't make sense. Two cases are random appeared and the ratio is about 1:1. I didn't change anything between running. |
| 147 | I tried using that 8x8 image chip (import as raster) and the same thing happened. |