Meehai commited on
Commit
dbe88f6
·
1 Parent(s): 0e15783

results for the other tasks

Browse files
Files changed (1) hide show
  1. README.md +22 -3
README.md CHANGED
@@ -207,8 +207,8 @@ all 60.456 73.261
207
  | [PHG-MAE-Distil](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 56.27 | 66.34 | 61.11 | 37.69 |
208
  | [PHG-MAE](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 55.32 | 63.80 | 63.18 | 38.98 |
209
  | [Mask2Former](https://openaccess.thecvf.com/content/CVPR2022/papers/Cheng_Masked-Attention_Mask_Transformer_for_Universal_Image_Segmentation_CVPR_2022_paper.pdf) | 216M | 53.97 | 63.37 | 60.55 | 37.98 |
210
- | [NGC(LR)](https://openaccess.thecvf.com/content/ICCV2023W/LIMIT/papers/Marcu_Self-Supervised_Hypergraphs_for_Learning_Multiple_World_Interpretations_ICCVW_2023_paper.pdf) | 32M | 40.76 | 46.51 | 45.59 | 30.17 |
211
- | [NGC-Distil](https://openaccess.thecvf.com/content/ICCV2023W/LIMIT/papers/Marcu_Self-Supervised_Hypergraphs_for_Learning_Multiple_World_Interpretations_ICCVW_2023_paper.pdf) | 32M | 40.31 | n/a | n/a | n/a |
212
  | [CShift](https://www.bmvc2021-virtualconference.com/assets/papers/0455.pdf){^1} | n/a | 39.67 | 46.27 | 43.67 | 29.09 |
213
  | [NGC](https://cdn.aaai.org/ojs/16283/16283-13-19777-1-2-20210518.pdf){^1} | 32M | 35.32 | 44.34 | 38.99 | 22.63 |
214
  | [SafeUAV](https://openaccess.thecvf.com/content_ECCVW_2018/papers/11130/Marcu_SafeUAV_Learning_to_estimate_depth_and_safe_landing_areas_for_ECCVW_2018_paper.pdf){^1} | 1.1M | 32.79 | n/a | n/a | n/a |
@@ -227,10 +227,29 @@ TODO: provide a simplified (non-weighted) metric which can be used on the entire
227
 
228
  #### Depth estimation
229
 
230
- TODO
 
 
 
 
 
 
231
 
232
  #### Camera normals estimation
233
 
 
 
 
 
 
 
 
 
 
 
 
 
 
234
  TODO
235
 
236
  Notes:
 
207
  | [PHG-MAE-Distil](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 56.27 | 66.34 | 61.11 | 37.69 |
208
  | [PHG-MAE](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 55.32 | 63.80 | 63.18 | 38.98 |
209
  | [Mask2Former](https://openaccess.thecvf.com/content/CVPR2022/papers/Cheng_Masked-Attention_Mask_Transformer_for_Universal_Image_Segmentation_CVPR_2022_paper.pdf) | 216M | 53.97 | 63.37 | 60.55 | 37.98 |
210
+ | [NHG(LR)](https://openaccess.thecvf.com/content/ICCV2023W/LIMIT/papers/Marcu_Self-Supervised_Hypergraphs_for_Learning_Multiple_World_Interpretations_ICCVW_2023_paper.pdf) | 32M | 40.76 | 46.51 | 45.59 | 30.17 |
211
+ | [NHG-Distil](https://openaccess.thecvf.com/content/ICCV2023W/LIMIT/papers/Marcu_Self-Supervised_Hypergraphs_for_Learning_Multiple_World_Interpretations_ICCVW_2023_paper.pdf) | 4.4M | 40.31 | n/a | n/a | n/a |
212
  | [CShift](https://www.bmvc2021-virtualconference.com/assets/papers/0455.pdf){^1} | n/a | 39.67 | 46.27 | 43.67 | 29.09 |
213
  | [NGC](https://cdn.aaai.org/ojs/16283/16283-13-19777-1-2-20210518.pdf){^1} | 32M | 35.32 | 44.34 | 38.99 | 22.63 |
214
  | [SafeUAV](https://openaccess.thecvf.com/content_ECCVW_2018/papers/11130/Marcu_SafeUAV_Learning_to_estimate_depth_and_safe_landing_areas_for_ECCVW_2018_paper.pdf){^1} | 1.1M | 32.79 | n/a | n/a | n/a |
 
227
 
228
  #### Depth estimation
229
 
230
+ Score is L1 error in meters.
231
+
232
+ | Model | #paramters | Score (↓) |
233
+ |:-|:-|:-|
234
+ | [PHG-MAE](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 16.13 |
235
+ | [NHG-Distil](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 16.64 |
236
+ | [SafeUAV](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 21.66 |
237
 
238
  #### Camera normals estimation
239
 
240
+ Score is L1 score (in angles) * 100.
241
+
242
+ | Model | #paramters | Score (↓) |
243
+ |:-|:-|:-|
244
+ | [NHG-Distil](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 11.71 |
245
+ | [PHG-MAE](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 12.35 |
246
+ | [SafeUAV](https://arxiv.org/pdf/2510.10068){^2} | 4.4M | 12.40 |
247
+
248
+
249
+ #### Multi-task results
250
+
251
+ One model, three tasks at once.
252
+
253
  TODO
254
 
255
  Notes: