Supervised Hashing with Latent Factor Model

REQUIREMENTS

pip install -r requirements.txt

1. pytorch >= 1.0
2. loguru

DATASETS

cifar10-gist.mat password: umb6

cifar-10_alexnet.t password: f1b7

nus-wide-tc21_alexnet.t password: vfeu

imagenet-tc100_alexnet.t password: 6w5i

USAGE

usage: run.py [-h] [--dataset DATASET] [--root ROOT]
              [--code-length CODE_LENGTH] [--num-samples NUM_SAMPLES]
              [--max-iter MAX_ITER] [--beta BETA] [--lamda LAMDA]
              [--topk TOPK]

LFH_PyTorch

optional arguments:
  -h, --help            show this help message and exit
  --dataset DATASET     Dataset name.
  --root ROOT           Path of dataset
  --code-length CODE_LENGTH
                        Binary hash code length.(default:
                        8,16,24,32,48,64,96,128)
  --num-samples NUM_SAMPLES
                        Number of samples.(default: 64)
  --max-iter MAX_ITER   Number of iterations.(default: 50)
  --beta BETA           Hyper-parameter.(default: 30)
  --lamda LAMDA         Hyper-parameter.(default: 1)
  --topk TOPK           Calculate top k data map.(default: all)

EXPERIMENTS

cifar10-gist dataset. Gist features, 1000 query images, 5000 training images. beta = 30, lamda = 1.

cifar-10-alexnet dataset. Alexnet features, 1000 query images, 5000 training images. beta=1, lamda = 50.

nus-wide-tc21-alexnet dataset. Alexnet features, top 21 classes, 2100 query images, 10500 training images. beta = 1, lamda = 50.

imagenet-tc100-alexnet dataset. Alexnet features, top 100 classes, 5000 query images, 10000 training images. beta = 10, lamda = 40.

Bits	8	16	24	32	48	64	96	128
cifar10-gist@ALL	0.2339	0.2866	0.2968	0.3258	0.3339	0.3285	0.3419	0.3551
cifar10-alexnet@ALL	0.2994	0.3892	0.4032	0.4047	0.4302	0.4324	0.4365	0.4480
nus-wide-tc21-alexnet@5000	0.6396	0.6615	0.6903	0.7131	0.7321	0.7423	0.7625	0.7619
imagenet-tc100-alexnet@1000	0.0943	0.1870	0.2999	0.3738	0.4327	0.4754	0.5199	0.5344