search

OpenDroneMap / ODMSemantic3D

An open photogrammetry dataset of classified 3D point clouds for automated semantic segmentation. CC BY-SA 4.0
Other
15 stars 7 forks source link

Add waterbury and fix #42

Closed HeDo88TH closed 1 year ago

HeDo88TH commented 1 year ago
u4gbot commented 1 year ago
sheffield_park_small Overall accuracy from 91.37% to 80.94% | ground | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 91.62% | 92.53% | 98.94% | 0.96 | | After | 86.62% | 87.53% | 98.81% | 0.93 | | Diff | -5.00% | -4.99% | -0.14% | -0.03 | | building | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 75.33% | 76.46% | 98.08% | 0.86 | | After | 4.72% | 4.73% | 98.17% | 0.09 | | Diff | -70.60% | -71.73% | 0.10% | -0.77 | | low_vegetation | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 80.03% | 94.96% | 83.58% | 0.89 | | After | 82.84% | 94.38% | 87.14% | 0.91 | | Diff | 2.81% | -0.58% | 3.56% | 0.02 | | human_made_object | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 12.76% | 86.17% | 13.03% | 0.23 | | After | 4.61% | 95.36% | 4.62% | 0.09 | | Diff | -8.15% | 9.19% | -8.41% | -0.14 |
brighton_beach_small Overall accuracy from 63.73% to 54.01% | ground | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 63.73% | 64.12% | 99.06% | 0.78 | | After | 53.45% | 54.50% | 96.53% | 0.70 | | Diff | -10.27% | -9.62% | -2.53% | -0.08 | | building | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 1.87% | 99.96% | 1.87% | 0.04 | | After | 0.80% | 48.84% | 0.80% | 0.02 | | Diff | -1.07% | -51.12% | -1.06% | -0.02 | | low_vegetation | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 55.99% | 58.07% | 94.01% | 0.72 | | After | 30.78% | 30.90% | 98.77% | 0.47 | | Diff | -25.22% | -27.17% | 4.76% | -0.25 | | human_made_object | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 0.00% | 0.00% | N/A | N/A | | After | 16.47% | 94.05% | 16.64% | 0.28 | | Diff | N/A | N/A | N/A | N/A |

Average (2 datasets)

Overall accuracy from 77.55% to 67.47%

ground Accuracy Recall Precision F1
Before 77.67% 78.32% 99.00% 0.87
After 70.04% 71.02% 97.67% 0.81
Diff -7.64% -7.30% -1.33% -0.05
building Accuracy Recall Precision F1
Before 38.60% 88.21% 49.97% 0.45
After 2.76% 26.78% 49.49% 0.05
Diff -35.84% -61.42% -0.48% -0.39
low_vegetation Accuracy Recall Precision F1
Before 68.01% 76.51% 88.79% 0.80
After 56.81% 62.64% 92.96% 0.69
Diff -11.20% -13.88% 4.16% -0.12
human_made_object Accuracy Recall Precision F1
Before 6.38% 43.08% 6.51% 0.11
After 10.54% 94.71% 10.63% 0.19
Diff 4.16% 51.62% 4.12% 0.07