MatConvNet is a MATLAB toolbox implementing Convolutional Neural Networks (CNNs) for computer vision applications. It is simple, efficient, and can run and learn state-of-the-art CNNs. Several example CNNs are included to classify and encode images. Please visit the homepage to know more.
In case of compilation issues, please read first the Installation and FAQ section before creating an GitHub issue. For general inquiries regarding network design and training related questions, please use the Discussion forum.
This repository is sharing a training and evaluation code for our papaer "Image Retrieval with Similar Object Detection and Local Similarity to Detected Objects" publihed in Pacific Rim International Conference on Artificial Intelligence (PRICAI), 2019. To run the code, first you place the pascal 2007 and Pascal 2012 datasets in the data folder and specify the path for root variable in the train.m code.
root = /path/to/data/
data => contain train and trest images for Pascal 2007 and 2012 datsets
Pretrained model for pascal 2007and 2012 and as well as INSTRE datasets are also provided in files
pascal 2007 => Model_trained_pascal_voc_2007.mat
pascal 2012 => Model_trained_pascal_voc_2012.mat
INSTRE => Model_trained_pascal_voc_INSTRE.mat
The evalution on Pascal 2007 and 2012 dataest for the recall of top-10 correct retrieval is given in the table below :
The qualitative results on Pascal datasets are shown in figure below. The "blue" bounding box is the region of interest in query image wheras the "yellow" bounding box shows the similar regions detected in database images.
