8000 GitHub - darolt/mask_rcnn: Pytorch 1.0 implementation of Mask RCNN
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

darolt/mask_rcnn

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

68 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

mask-rcnn

This is a Pytorch 1.0 implementation of Mask R-CNN that is based on Matterport's Mask_RCNN[1] and this[2]. Matterport's repository is an implementation on Keras and TensorFlow.

The main improvements from [2] are:

  • Pytorch 1.0
  • most numpy computations were ported to pytorch (for GPU speed)
  • supports batchsize > 1
  • some bugs were fixed in the translation process
  • code refactor
  • NMS speed-up

Currently, it works with Kaggle's 2018 Data Science Bowl dataset (the result on 1st phase testset is 0.27).

to train the network use: python samples/nuclei.py train --dataset=path_to_dataset --model=coco

to detect use: python samples/nuclei.py submit --dataset=path_to_dataset --model=path_to_trained_model

to check Kaggle's 2018 Databowl metric on a dataset use: python samples/nuclei.py metric --dataset=path_to_dataset --model=path_to_trained_model

for installation instructions, just export mrcnn directory to PYTHONPATH and run: python setup.py install

About

Pytorch 1.0 implementation of Mask RCNN

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published
0