Introducing
Your new presentation assistant.
Refine, enhance, and tailor your content, source relevant images, and edit visuals quicker than ever before.
Trending searches
http://github.com/tkorting/youtube/
Individual decision trees can be used to retrieve class uncertainty
shadow Vs water
| case | cloud | free | shadow | water | entropy |
| 1 | 100 | 0 | 0 | 0 | 0.000 |
| 2 | 0 | 0 | 25 | 75 | 0.562 |
| 3 | 0 | 50 | 0 | 50 | 0.693 |
| 4 | 25 | 25 | 0 | 50 | 1.039 |
transition cloud-free
Random Forest for image classification
http://youtube.com/tkorting
Thales Sehn Körting
Class prediction using Random Forests
Random Forests for image classification
1
[[ 54 49 49 48 48 48 48 48 51 51]
[ 45 46 46 43 43 45 45 45 44 44]
[ 45 46 46 43 43 45 45 45 44 44]
[ 61 90 90 82 82 48 42 42 47 47]
[138 139 139 122 122 78 47 47 46 46]
[138 139 139 122 122 78 47 47 46 46]
[144 144 144 140 140 124 70 70 48 48]
[144 144 144 140 140 124 70 70 48 48]
[146 144 144 142 142 145 80 80 51 51]
[146 144 144 142 142 145 80 80 51 51]]
[[ 22 20 20 19 19 19 19 19 20 20]
[ 9 11 11 7 7 16 16 16 17 17]
[ 9 11 11 7 7 16 16 16 17 17]
[ 25 54 54 46 46 20 14 14 20 20]
[ 86 90 90 74 74 41 10 10 22 22]
[ 86 90 90 74 74 41 10 10 22 22]
[ 92 94 94 90 90 86 33 33 25 25]
[ 92 94 94 90 90 86 33 33 25 25]
[ 93 93 93 91 91 108 43 43 22 22]
[ 93 93 93 91 91 108 43 43 22 22]]
Sample 1 - bare soil [94, 144, 206]
Sample 2 - vegetation [16, 45, 21]
[[ 20 23 23 22 22 20 20 20 21 21]
[ 40 41 41 37 37 21 21 21 18 18]
[ 40 41 41 37 37 21 21 21 18 18]
[ 60 96 96 86 86 24 19 19 21 21]
[210 199 199 173 173 74 38 38 22 22]
[210 199 199 173 173 74 38 38 22 22]
[220 206 206 200 200 129 65 65 24 24]
[220 206 206 200 200 129 65 65 24 24]
[220 211 211 207 207 167 84 84 25 25]
[220 211 211 207 207 167 84 84 25 25]]
Sample N - class name ...
1
in this example:
- 4 input features
- 4 classes (~31000 samples)
- 4 trees (in the forest)
input
- blue
- green
- red
- nir
output:
- cloud
- free
- shadow
- water
2
3
Majority vote = cloud (3 votes)