else:
loss_midn = midn_loss.item()
loss_oicr = oicr_loss.item()
loss_oicr1 = oc1.item()
loss_oicr2 = oc2.item()
loss_oicr3 = oc3.item()
fg_cnt = torch.sum(rois_label.data.ne(0))
bg_cnt = rois_label.data.numel() - fg_cnt
record_module = OICR
a = torch.max(im_rois[:, :, 1]).data
b = torch.max(im_rois[:, :, 2]).data
c = torch.max(im_rois[:, :, 3]).data
d = torch.max(im_rois[:, :, 4]).data
logger.log("[session %d][epoch %2d][iter %4d/%4d] loss: %.4f, lr: %.2e" \
% (args.session, epoch, step, iters_per_epoch, loss_temp, lr))
logger.log("\t\t\tfg/bg=(%d/%d), time cost: %f" %
(fg_cnt, bg_cnt, end - start))
#print("\t\t\tmidn : %.4f, oicr : %.4f" % (loss_midn, loss_oicr))
logger.log("\t\t\tmidn : %.4f, oicr : %.4f" %
(loss_midn_temp, loss_oicr_temp))
logger.log("\t\t\tdata loading : %.4f" % (loading_time))
loading_time = 0
print(c, d, im_data.shape)
logger.log("Logging to {}".format(log_dir))
# end batch
# logging
logger.record_tabular('loss', loss_temp)
logger.record_tabular('midn_loss', loss_midn_temp)
logger.record_tabular('oicr_loss', loss_oicr_temp)
for j in range(imdb.num_classes):
for i in range(num_images):
inds = np.where(all_boxes[j][i][:, -1] > thresh[j])[0]
all_boxes[j][i] = all_boxes[j][i][inds, :]
model_save_gt = {'images': images_real, 'gt': gt}
sio.savemat('{}_gt.mat'.format(imdb.name), model_save_gt)
det_file = os.path.join(output_dir_map, 'detections.pkl')
with open(det_file, 'wb') as f:
pickle.dump(all_boxes, f, pickle.HIGHEST_PROTOCOL)
det_file = os.path.join(output_dir_corloc, 'discovery.pkl')
with open(det_file, 'wb') as f:
pickle.dump(all_boxes_corloc, f, pickle.HIGHEST_PROTOCOL)
# due to memory issue
# det_file_all = os.path.join(output_dir, 'detections_all.pkl')
# results_all = {'scores_all' : scores_all, 'boxes_all' : boxes_all}
# with open(det_file_all, 'wb') as f:
# pickle.dump(results_all, f, pickle.HIGHEST_PROTOCOL)
print('Applying NMS to all detections')
nms_dets = apply_nms(all_boxes, cfg.TEST.NMS)
logger.log('Evaluating detections')
imdb.evaluate_detections(nms_dets, output_dir_map)
logger.log('Evaluating CorLoc')
imdb.evaluate_discovery(all_boxes_corloc, output_dir_corloc)