-
Download the project from the git repository.
-
Open the terminal
-
Install the dependencies of the caffe found on following link. http://caffe.berkeleyvision.org/installation.html
-
Install all required dependencies for executing Python 2.7: i. Open terminal at CBIR/python and execute following command for req in $(cat requirements.txt); do pip install $req; done
-
Execute following commands: i. make all -j8 ii. make test -j8 iii. make runtest iv. make pycaffe v. ./prepare.sh vi. cd /examples/cbir-cifar10/ chmod 777 train_CPU_48 train_GPU_48 ./train_CPU_48 -- for training with CPU ./train_GPU_48 -- for training with GPU
-
Add the CBIR/python to your python path.
-
Modify the model with newly created model in step 5.vi. and execute the RunPythonCIfarGpu/Cpu.py - This will train and test the code.
-
If you want to manually download the dataset, we have kept it at : https://www.dropbox.com/l/scl/AAASnbNHfs9b9BF5kPMTplWVJc45X9e3dIU i. Extract the dataset and keep it at CBIR/examples/cbir-cifar10/dataset
-
The trained model for GPU is kept at this location: https://www.dropbox.com/s/dlak33m5icvdg1o/CBIR_GPU.caffemodel?dl=0
-
The trained model for CPU is kept at this location: https://www.dropbox.com/s/iixuym610fntld6/CBIR_CPU.caffemodel?dl=0