update README

README.md

@@ -4,7 +4,8 @@ By Xuanyi Dong, Liang Zheng, Fan Ma, Yi Yang, Deyu Meng
 
 ### Introduction
 
-This framework aims to solve the few-example object detection problem, in which there are only a few images with bounding box annotations per object class and a large number of images without annotations. The framework combines self-paced learning and multi-modal learning, and we call it Multi-modal Self-Paced Learning for Detection (MSPLD). The code is modified from [R-FCN](https://github.com/daijifeng001/R-FCN).
+This framework aims to solve the few-example object detection problem, in which there are only a few images with bounding box annotations per object class and a large number of images without annotations. The framework combines self-paced learning and multi-modal learning, and we call it Multi-modal Self-Paced Learning for Detection (MSPLD). The code is modified from [R-FCN](https://github.com/daijifeng001/R-FCN) and [Faster RCNN](https://github.com/shaoqingren/faster_rcnn).
+Note that few-example object detection is a kind of semi-supervised object detection.
 
 MSPLD was accepted to [TPAMI 2018](https://ieeexplore.ieee.org/document/8374906/).
 

external/README.md

@@ -0,0 +1,11 @@
+# External Library
+
+## Caffe
+Caffe is provided the `caffe` directory.
+
+## Information
+In our ablative studies, we use the Non-overlapping ImageNet-Voc and Non-overlapping ImageNet-COCO to pre-train the classification models.
+The Non-overlapping ImageNet-Voc is the ImageNet-1k data without the PASCAL VOC 20 classes, and the Non-overlapping ImageNet-COCO is similar to Non-overlapping ImageNet-VOC.
+The `information` directory contains the classes in the Non-overlapping ImageNet-Voc and the Non-overlapping ImageNet-COCO.
+
+See the list file : `ImageNet-1000.coco` and `ImageNet-1000.pvoc`. As these two lists are manually labeled by myself, there might be some mistakes.

functions/weakly/online/prepare_online_voc.m

@@ -1,22 +1,11 @@
 function res_name = prepare_online_voc(cls, boxes, imdb, cache_name, suffix)
-% res = imdb_eval_voc(cls, boxes, imdb, suffix)
-%   Use the VOCdevkit to evaluate detections specified in boxes
-%   for class cls against the ground-truth boxes in the image
-%   database imdb. Results files are saved with an optional
-%   suffix.
+% --------------------------------------------------------
+% MSPLD implementation
+% Modified from MATLAB Faster R-CNN (https://github.com/shaoqingren/faster_rcnn)
+% Copyright (c) 2018, Xuanyi Dong
+% Licensed under The MIT License [see LICENSE for details]
+% --------------------------------------------------------
 
-% AUTORIGHTS
-% ---------------------------------------------------------
-% Copyright (c) 2014, Ross Girshick
-% 
-% This file is part of the R-CNN code and is available 
-% under the terms of the Simplified BSD License provided in 
-% LICENSE. Please retain this notice and LICENSE if you use 
-% this file (or any portion of it) in your project.
-% ---------------------------------------------------------
-
-% Add a random string ("salt") to the end of the results file name
-% to prevent concurrent evaluations from clobbering each other
 use_res_salt = true;
 % comp4 because we use outside data (ILSVRC2012)
 comp_id = 'comp4';

functions/weakly/online/weakly_online_test.m

@@ -1,3 +1,8 @@
+%###########################################################################
+%## Copyright (c) 2018-present, Xuanyi Dong                              ###
+%## Few-Example Object Detection with Model Communication                ###
+%## IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018 ###
+%###########################################################################
 function res = weakly_online_test(confs, imdb, roidb, varargin)
     ip = inputParser;
     ip.addRequired('confs',                             @iscell);

functions/weakly/supervise/weakly_train_mode.m

@@ -1,3 +1,8 @@
+%###########################################################################
+%## Copyright (c) 2018-present, Xuanyi Dong                              ###
+%## Few-Example Object Detection with Model Communication                ###
+%## IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018 ###
+%###########################################################################
 function mode = weakly_train_mode( conf )
     if ( conf.regression ) 
         if (conf.fast_rcnn)

functions/weakly/utils/rfcn_bbox_transform.m

@@ -1,3 +1,9 @@
+% --------------------------------------------------------
+% MSPLD implementation
+% Modified from MATLAB Faster R-CNN (https://github.com/shaoqingren/faster_rcnn)
+% Copyright (c) 2018, Xuanyi Dong
+% Licensed under The MIT License [see LICENSE for details]
+% --------------------------------------------------------
 function [regression_label] = rfcn_bbox_transform(ex_boxes, gt_boxes)
 
     ex_widths = ex_boxes(:, 3) - ex_boxes(:, 1) + 1;

functions/weakly/utils/rfcn_bbox_transform_inv.m

@@ -1,3 +1,9 @@
+% --------------------------------------------------------
+% MSPLD implementation
+% Modified from MATLAB Faster R-CNN (https://github.com/shaoqingren/faster_rcnn)
+% Copyright (c) 2018, Xuanyi Dong
+% Licensed under The MIT License [see LICENSE for details]
+% --------------------------------------------------------
 function [pred_boxes] = rfcn_bbox_transform_inv(boxes, box_deltas)
 
     src_w = double(boxes(:, 3) - boxes(:, 1) + 1);

functions/weakly/weakly_co_train_v3.m

@@ -123,8 +123,6 @@
     pre_keeps = zeros(numel(unsupervise_ids), numel(models));
 
     % rng_seed 5; per_sample 3;
-    %Init_Per_Select = [40, 12, 10, 20, 20, 10, 50, 25, 15, 10,...
-    %                   20, 15, 15, 30, 15, 15, 15, 20, 40, 35];
     Init_Per_Select = [40, 12, 10, 20, 20, 10, 50, 25, 15, 10,...
                        20, 15, 15, 30, 15, 15, 15, 20, 40, 35];
 %% Start Training