Awesome

pytorch-NetVlad

Implementation of NetVlad in PyTorch, including code for training the model on the Pittsburgh dataset.

Reproducing the paper

Below are the result as compared to the results in third row in the right column of Table 1:

Running main.py with train mode and default settings should give similar scores to the ones shown above. Additionally, the model state for the above run is available here: https://drive.google.com/open?id=17luTjZFCX639guSVy00OUtzfTQo4AMF2

Using this checkpoint and the following command you can obtain the results shown above:

python main.py --mode=test --split=val --resume=vgg16_netvlad_checkpoint/

Setup

Dependencies

1. PyTorch (at least v0.4.0)
2. Faiss
3. scipy
   • numpy
   • sklearn
   • h5py
4. tensorboardX

Data

Running this code requires a copy of the Pittsburgh 250k (available here), and the dataset specifications for the Pittsburgh dataset (available here). pittsburgh.py contains a hardcoded path to a directory, where the code expects directories 000 to 010 with the various Pittsburth database images, a directory queries_real with subdirectories 000 to 010 with the query images, and a directory datasets with the dataset specifications (.mat files).

Usage

main.py contains the majority of the code, and has three different modes (train, test, cluster) which we'll discuss in mode detail below.

Train

In order to initialise the NetVlad layer it is necessary to first run main.py with the correct settings and --mode=cluster. After which a model can be trained using (the following default flags):

python main.py --mode=train --arch=vgg16 --pooling=netvlad --num_clusters=64

The commandline args, the tensorboard data, and the model state will all be saved to opt.runsPath, which subsequently can be used for testing, or to resuming training.

For more information on all commandline arguments run:

python main.py --help

Test

To test a previously trained model on the Pittsburgh 30k testset (replace directory with correct dir for your case):

python main.py --mode=test --resume=runsPath/Nov19_12-00-00_vgg16_netvlad --split=test

The commandline arguments for training were saved, so we shouldnt need to specify them for testing. Additionally, to obtain the 'off the shelf' performance we can also omit the resume directory:

python main.py --mode=test

Cluster

In order to initialise the NetVlad layer we need to first sample from the data and obtain opt.num_clusters centroids. This step is necessary for each configuration of the network and for each dataset. To cluster simply run

python main.py --mode=cluster --arch=vgg16 --pooling=netvlad --num_clusters=64

with the correct values for any additional commandline arguments.