Live-Eye--Large-Scale-Object-Recognition

This project uses deep machine learning to perform object recognition at large scale.

We will be using state of the art deep machine learning tools and technologies to perform object recognition on more than 2,00,000 images. The dataset used here is from imagenet and Microsoft COCO challenge 2015.

Install Caffee

http://caffe.berkeleyvision.org/install_apt.html

sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler

sudo apt-get install --no-install-recommends libboost-all-dev

Install CUDA :

sudo dpkg -i cuda-repo-ubuntu1404-7-5-local_7.5-18_amd64.deb

sudo apt-get update

sudo apt-get install cuda

Install BLAS - Atlas is default for caffe :

sudo apt-get install libatlas-base-dev

Install the python-dev package

To have the Python headers for building the pycaffe interface

sudo apt-get install python-dev

Install cudnn

LINUX

cd <installpath>
export LD_LIBRARY_PATH=`pwd`:$LD_LIBRARY_PATH

Add <installpath> to your build and link process by adding -I<installpath> to your compile
line and -L<installpath> -lcudnn to your link line.

Other dependencies.

sudo cp include/cudnn.h /usr/local/cuda-7.5/include/

sudo cp lib64/libcudnn* /usr/local/cuda-7.5/lib64/

Before training.

export LD_LIBRARY_PATH=/usr/local/cuda-7.5/lib64:$LD_LIBRARY_PATH

Training command

tools/train_net.py --gpu 0 --solver ./models/solvers/solver.prototxt --weights data/imagenet_models/VGG16.v2.caffemodel --imdb coco_2014_train --iters 500 --cfg ./experiments/cfgs/faster_rcnn_end2end.yml

Testing command

./tools/test_net.py --gpu 0 --def ./models/coco/solvers/test.prototxt --net ./models/bigdata_coco_final.caffemodel --imdb coco_2015_test --cfg experiments/cfgs/faster_rcnn_end2end.yml