This repository was created by combining Cartucho's mAP with

the R2CNN-Plus-Plus_Tensorflow/blob/master/libs/box_utils/iou_rotate.py store in DetectionTeamUCAS.

* Cartucho's mAP

* DetectionTeamUCAS's R2CNN-Plus_Plus_Tensorflow

RBBox mAP(Rotated Bounding Box mean Average Precision)

• rotated output
  
• horizontal output
  
• rotated ground truth on horizontal predict output

HOW TO USE

You should have to build cython file if you want to using GPU on calculating IoU. You don't have to if you using cpu.

$ python setup.py build_ext --inplace

use_gpu

$ python main.py --gpu gpu

use_cpu

$ python main.py

IF YOU USING ON GOOGLE COLAB, MUST HAVE TO

$ python main.py -na --gpu (cpu,gpu)

Animator does not work on colab.

Bounding Box Mode example

Box mode has default value as 'rotated'. So if you want measure on horizontal prediction, you should have to give each box mode

$ python main.py --gt-box-mode horizontal --pred-box-mode horizontal

If you want other information of mean Average Precision, please visit Cartucho's mAP repository.

There has many Reference like IoU and other things.

Authors:

• chromatices - Please give me your feedback: to. [email protected]