Fine-Grained Visual Classification via Progressive Multi-Granularity Training of Jigsaw Patches

This respository is implemnet for latest Version for "Fine-Grained Visual Classification via Progressive Multi-Granularity Training of Jigsaw Patches". This respository don't cause an ERROR

This Project training 3 files at once as a Mixture of Granularity-Specific Experts Convolutional Neural Net and stores Experimental results. additionaly use wandb.

For paper implementations, see the section "Papers and projects".

Framework

Setup the enviroment for evaluation

$cd PMG
$sh test.sh 

Default

Default inference. in python main.py --seed 0 --dataset cub --imgsize 550 --crop 448 --model resnet50 --epochs 300 --batchsize 16 --gpu_ids 0,1.

Datasets

We implemented to load dataset, so It'll work if you just run it. but, CUB dataset needs to be download.

• CUB-200-2011
• Stanford Cars
• FGVCaircraft

File Structure

├── dataset
│   ├── data
│   │     ├── aircraft
│   │     ├── stanfordcar
│   │     └── cubbird
│   ├── __init__.py
│   ├── augmentataion.py
│   └── etc..
│
├── models
│   ├── __init__.py
│   ├── base.py   
│   ├── grad_cam.py
│   └── local_cam.py
│
├── trainer
│   ├── __init__.py
│   ├── train.py   
│   └── infer.py
│   
├── main.py 
└── etc..

Papers and projects

| Name | Location | Comment | | :------------------------------------------------------------ | :-------------------------------------------------------------- | :------------- | | Fine-Grained Visual Classification via Progressive Multi-Granularity Training of Jigsaw Patches | link | ECCV 2020 | | Learning a Mixture of Granularity-Specific Experts for Fine-Grained Categorization | link | ICCV 2019 |

How to cite

@article = {
    title = {Fine-Grained Visual Classification via Progressive Multi-Granularity Training of Jigsaw Patches},
    author = {Wongi Park},
    journal = {GitHub},
    url = {https://github.com/kalelpark/Latest_Progressive-Multi-Granularity-Training-of-Jigsaw-Patches},
    year = {2022},
}