Skip to content

Latest commit

 

History

History
24 lines (19 loc) · 1005 Bytes

RESULTS.md

File metadata and controls

24 lines (19 loc) · 1005 Bytes
Raw

Results

We provide Faster R-CNN and RetinaNet results on Tiny-DOTA datasets, and a script for drawing number of clicks (NoC) graph.

Draw NoC Curve

python draw_noc_graph.py

Note that you can obtain new models' NoC curve.

  1. Add new model results as XXX.json files in the Tiny-DOTA-FasterRCNN or Tiny-DOTA-RetinaNet folder.
  2. Change code in the draw_noc_graph.py as follow:

files = ['Ours.json', 'Late_Fusion.json', 'Early_Fusion.json', 'Passthrough.json', 'Faster_R-CNN.json']

Faster R-CNN

Faster R-CNN Results

Each json file represents our model and baseline results which included in average of mAP and standard deviation calculated by five times inference results.

RetinaNet

RetinaNet Results

Each json file represents our model and baseline results which included in average of mAP and standard deviation calculated by five times inference results.