example.lua

require 'ROIDetection'
require 'config'
require 'pascal_voc'
require 'torch'
require 'image'
require 'utils'
require 'data_layer'


-- example for ROIDetection
-- year = '2007'
-- data_type = 'train'
-- gt_bboxes, gt_classes = pascal_voc.get_gt_bboxes(year, data_type)
gt_bboxes, gt_classes = pascal_voc.get_gt_bboxes()
-- ss_rois = pascal_voc.get_ss_rois(year, data_type)
ss_rois = pascal_voc.get_ss_rois()
means, stds, roi_set, bbox_targets, max_overlaps = 
    ROIDetection.add_bbox_regression_targets(ss_rois, gt_bboxes, gt_classes)
images = pascal_voc.load_image_set_index()

------ test data_layer -------------------
--[=[
data_layer.set_roidb(images, roi_set, bbox_targets, max_overlaps)
while true do
    local im_blob, rois_blob, bbox_targets_blob, labels_blob = data_layer.get_next_minibatch()
    print(#im_blob, #rois_blob, #bbox_targets_blob, #labels_blob)
    io.read()
    print('===========================================================================\n\n')
end


------ test nms --------------------------
math.randomseed(tostring(os.time()):reverse():sub(1, 6))
len = gt_bboxes[2059]:size(1)
scores = torch.Tensor(len):zero()
scores:apply(function()
    return math.random()
end)
temp = torch.Tensor(len, 5)
temp[{{}, {1}}] = scores
temp[{{}, {2, 5}}] = gt_bboxes[2059]
print(temp)
pick = utils.nms(temp, 0.0)
print(pick)
]=]