Nicholas Westlake

Results 10 comments of Nicholas Westlake

Unknown layer_type ROIPooling

This sounds to me like it's running the wrong version of Caffe, i.e. not the one required for and included with fast-rcnn which includes the appropriate layer. I suspect running...

Train and Test R-CNN on Another Dataset

@leejiajun Try applying the changes proposed here: https://github.com/rbgirshick/fast-rcnn/pull/102 e.g. git fetch [email protected]:nw362/fast-rcnn.git negativeImages:negativeImages git merge negativeImages (I am interesting in the pull request too, so I might later try merging...

Train and Test R-CNN on Another Dataset

This looks to be code you have written (amph.py)? Hard to say without seeing both the code and the annotation file.

Train and Test R-CNN on Another Dataset

One ROI at a time?

Train and Test R-CNN on Another Dataset

The convolutional layers of the CNN are calculated on the whole image (resized before hand). The final polling layer pools over only the ROI and so the CNN must run...

Train and Test R-CNN on Another Dataset

It should happen already "lib/fast_rcnn/test.py": _get_blobs will take in one image but multiple ROIs. In im_detect(net, im, boxes), the CNN will be called forward only once for an image and...

Train and Test R-CNN on Another Dataset

Yes, e.g. the xml format used for annotations.

Train and Test R-CNN on Another Dataset

No problem at all. This is the case for many data sets.

How to manually add negative samples for training to improve precision

Set TRAIN.BG_THRESH_LO to 0.0?

How to use negative examples when training

Please see https://github.com/rbgirshick/fast-rcnn/pull/102 NB by default none of the ROIs in negative image will be used, however you can set TRAIN.BG_THRESH_LO to 0.0.