search

rbgirshick / py-faster-rcnn

Faster R-CNN (Python implementation) -- see https://github.com/ShaoqingRen/faster_rcnn for the official MATLAB version
Other
8.13k stars 4.11k forks source link

Incorporating images w/ no annotated objects #402

Open bkj opened 8 years ago

bkj commented 8 years ago

Hi all,

Does anyone know whether it's possible to train py-faster-rcnn models using images that don't have any annotations -- eg, they're all "background"? Even better, is there a way to provide explicit "negative annotations" -- eg, annotate things that look roughly like class_X but are actually background?

I'm planning to look through the code to figure it out, but wondered if anyone knew off the top of their heads.

Thanks Ben

marifnst commented 8 years ago

Hi @bkj

I have implemented faster r cnn for texture detection using dataset https://www.robots.ox.ac.uk/~vgg/data/dtd/. Faster r cnn give good AP with same data for training and testing.

Any comment from other ?

thanks marifnst

sbagchi12 commented 7 years ago

@marifnst It will be great if you can share how did you organize the dtd dataset as an input to the faster RCNN framework? Did you give the entire image as the bounding box to represent the texture as an object? Thanks in advance