Classification is a Strong Baseline for Deep Metric Learning (BMVC '19)

Andrew Zhai, Hao-Yu Wu

Paper (https://arxiv.org/abs/1811.12649)

This repo contains the source code for our paper (WIP)

Setup Repo

git clone https://github.com/azgo14/classification_metric_learning.git

The repo assumes that all data files exist under /data1 directory locally as that local directly will be mounted ot /data1 in the container.

Running Commands

We provide a simple utility to make running commands in a docker container easier. This tool will automatically download the expected Docker image, mount the expected directories, and make running commands simple. To use the command, add scripts/bin/pdoc to your PATH variable as so:

export PATH=$PATH:<PATH_TO_REPO>/scripts/bin

Example

You can then see that commands such as pdoc nvidia-smi will run nvidia-smi inside the docker container.

Build Docker

To rebuild the docker image:

1. Initialize all submodules recursively via:

git submodule update --init --recursive

2. Build the image with

./docker/docker-build.sh ./docker/Dockerfile

Datasets

Download the datasets with the following scripts. We assume data will live in the /data1 directory throughput our code

CUB

./scripts/get_cub200_dataset.sh

CARS

./scripts/get_cars196_dataset.sh

Stanford Online Products

./scripts/get_stanford_products_dataset.sh

In-Shop

Manual download raw data from http://mmlab.ie.cuhk.edu.hk/projects/DeepFashion/InShopRetrieval.html

Expected the following raw data files to exist /data1/data/inshop/img.zip /data1/data/inshop/list_eval_partition.txt

./scripts/get_inshop_dataset.sh

Reproduction

CUB

./scripts/bin/pdoc CUDA_VISIBLE_DEVICES=0 CUDA_DEVICE_ORDER=PCI_BUS_ID python metric_learning/train_classification.py --dataset Cub200 --dim 2048 --model_name resnet50 --epochs_per_step 15 --num_steps 2 --test_every_n_epochs 5 --lr 0.001 --lr_mult 1 --class_balancing --images_per_class 25 --batch_size 75

(May differ slightly because of random seed)
Raw Features: R@1, R@2, R@4, R@8: 65.36 & 76.76 & 85.42 & 91.51
Binary Features: R@1, R@2, R@4, R@8: 63.67 & 75.37 & 84.54 & 90.99

CARS

./scripts/bin/pdoc CUDA_VISIBLE_DEVICES=1 CUDA_DEVICE_ORDER=PCI_BUS_ID python metric_learning/train_classification.py --dataset Cars196 --dim 2048 --model_name resnet50 --epochs_per_step 15 --num_steps 2 --test_every_n_epochs 5 --lr 0.01 --lr_mult 1 --class_balancing --images_per_class 25 --batch_size 75

(May differ slightly because of random seed)
Raw Features: R@1, R@2, R@4, R@8: 89.50 & 94.18 & 96.84 & 98.41
Binary Features: R@1, R@2, R@4, R@8: 89.29 & 93.95 & 96.61 & 98.14

Stanford Online Products

./scripts/bin/pdoc CUDA_VISIBLE_DEVICES=1 CUDA_DEVICE_ORDER=PCI_BUS_ID python metric_learning/train_classification.py --dataset StanfordOnlineProducts --dim 2048 --model_name resnet50 --epochs_per_step 15 --num_steps 2 --test_every_n_epochs 5 --lr 0.01 --lr_mult 1 --class_balancing --images_per_class 5 --batch_size 75

(May differ slightly because of random seed)
Raw Features: R@1, R@10, R@100, R@1000: 79.55 & 91.54 & 96.66 & 98.95
Binary Features: R@1, R@10, R@100, R@1000: 78.03 & 90.71 & 96.24 & 98.72

In-Shop

./scripts/bin/pdoc CUDA_VISIBLE_DEVICES=5 CUDA_DEVICE_ORDER=PCI_BUS_ID python metric_learning/train_classification.py --dataset InShop --dim 2048 --model_name resnet50 --epochs_per_step 15 --num_steps 2 --test_every_n_epochs 5 --lr 0.01 --lr_mult 1 --class_balancing --images_per_class 5 --batch_size 75

(May differ slightly because of random seed)
Raw Features: R@1, R@10, R@20, R@30, R@40, R@50: 89.35 & 97.81 & 98.61 & 98.87 & 99.05 & 99.13
Binary Features: R@1, R@10, R@20, R@30, R@40, R@50: 88.76 & 97.65 & 98.47 & 98.73 & 98.94 & 99.05

References / re-implementations

• The Computer Vision Best Practices repository: 02_state_of_the_art.ipynb