lzy_HCKC: A Python repository from fhqxa

Hierarchical convolutional neural network with knowledge complementation for long-tailed classification

HONG ZHAO*, ZHENGYU LI, WENWEI HE, and Yan Zhao

Main requirements

torch == 1.0.1
torchvision == 0.2.2_post3
tensorboardX == 1.8
Python 3

Environmental settings

The code is developed using the PyTorch framework. We conduct experiments on a single NVIDIA GeForce RTX 2080 Ti GPU. The CUDA nad CUDNN version is 9.0 and 7.1.3 respectively. Other platforms or GPU cards are not fully tested.

Usage

# To train long-tailed CIFAR-10 with imbalanced ratio of 50:
python main/train.py  --cfg configs/cifar10.yaml     

# To validate with the best model:
python main/valid.py  --cfg configs/cifar10.yaml

# To debug with CPU mode:
python main/train.py  --cfg configs/cifar10.yaml   CPU_MODE True

You can change the experimental setting by simply modifying the parameter in the yaml file.

Data format

The annotation of a dataset is a dict consisting of two field: annotations and num_classes. The field annotations is a list of dict with image_id, fpath, im_height, im_width and category_id.

Pretrain models for iNaturalist

We provide the BBN pretrain models of both 1x scheduler and 2x scheduler for iNaturalist 2018 and iNaturalist 2017.

iNaturalist 2018: Baidu Cloud, Google Drive

iNaturalist 2017: Baidu Cloud, Google Drive

The experimental setup was as follows:

python main.py --dataset cifar10 -a resnet32 --num_classes 10 --imbanlance_rate 0.01 --beta 0.5 --lr 0.01 --epochs 200 -b 64 --momentum 0.9 --weight_decay 5e-3 --resample_weighting 0.0 --label_weighting 1.2 --contrast_weight 4

Overview

figure_3method1.pdf

Preparing Datasets

Download the datasets CIFAR-10, CIFAR-100, ImageNet, and iNaturalist18 to GLMC-2023/data. The directory should look like

GLMC-2023/data
├── CIFAR-100-python
├── CIFAR-10-batches-py
├── ImageNet
|   └── train
|   └── val
├── train_val2018
└── data_txt
    └── ImageNet_LT_val.txt
    └── ImageNet_LT_train.txt
    └── iNaturalist18_train.txt
    └── iNaturalist18_val.txt

Training

for CIFAR-10-LT

python main.py --dataset cifar10 -a resnet32 --num_classes 10 --imbanlance_rate 0.01 --beta 0.5 --lr 0.01 --epochs 200 -b 64 --momentum 0.9 --weight_decay 5e-3 --resample_weighting 0.0 --label_weighting 1.2 --contrast_weight 1

python main.py --dataset cifar10 -a resnet32 --num_classes 10 --imbanlance_rate 0.02 --beta 0.5 --lr 0.01 --epochs 200 -b 64 --momentum 0.9 --weight_decay 5e-3 --resample_weighting 0.0 --label_weighting 1.2 --contrast_weight 1

python main.py --dataset cifar10 -a resnet32 --num_classes 10 --imbanlance_rate 0.1 --beta 0.5 --lr 0.01 --epochs 200 -b 64 --momentum 0.9 --weight_decay 5e-3 --resample_weighting 0.2 --label_weighting 1  --contrast_weight 2

for CIFAR-100-LT

python main.py --dataset cifar100 -a resnet32 --num_classes 100 --imbanlance_rate 0.01 --beta 0.5 --lr 0.01 --epochs 200 -b 64 --momentum 0.9 --weight_decay 5e-3 --resample_weighting 0.0 --label_weighting 1.2  --contrast_weight 4

python main.py --dataset cifar100 -a resnet32 --num_classes 100 --imbanlance_rate 0.02 --beta 0.5 --lr 0.01 --epochs 200 -b 64 --momentum 0.9 --weight_decay 5e-3 --resample_weighting 0.2  --label_weighting 1.2  --contrast_weight 6

python main.py --dataset cifar100 -a resnet32 --num_classes 100 --imbanlance_rate 0.1 --beta 0.5 --lr 0.01 --epochs 200 -b 64 --momentum 0.9 --weight_decay 5e-3 --resample_weighting 0.2  --label_weighting 1.2  --contrast_weight 4

fhqxa/lzy_HCKC