mxbi / kaggle-kuzushiji-recognition

Kaggle Kuzushiji Recognition Solution

kaggle-kuzushiji-recognition

Project structure

├── code
│   ├── configs
│   ├── crop_det.py
│   ├── make_submission.py
│   ├── mmdetection
│   ├── prepare_det.py
│   └── work_dirs
├── data
│   ├── dtrainval.pkl
│   ├── train_images
│   ├── test_images
│   ├── train_crops
│   └── ...
├── download
│   ├── test_images.zip
│   ├── train_images.zip
│   └── ...
└── submits
    ├── submit001.csv
    └── ...

Requirements

• 2x 1080 Ti
• mmdetection with a few modifications (e.g. mmdetection/mmdet/datasets/pipelines/test_aug.py)
• requirements.txt

Overview

The solution is straightfoward.

Cascade R-CNN with:

• Strong backbones
• Multi-scale train&test

Due to limited GPU memory, models were trained on 1024x1024 crops and tested on full images(with a max size limit).

LB score 0.935 with:

• HRNet w32
• train scales 512~768
• test scales [0.5, 0.625, 0.75]

LB score 0.946 with:

• HRNet w32
• train scales 768~1280
• test scales [0.75, 0.875, 1.0, 1.125, 1.25]

Ensembling HRNet_w32 and HRNet_w48 results -> 0.950.

Run

Install mmdetection following ./mmdetection/docs/INSTALL.md

Install packages following requirements.txt

Preprocess

python prepare_det.py
python crop_det.py

Train

./mmdetection/tools/dist_train.sh configs/hr32.py 2 --seed 0

Test

./mmdetection/tools/dist_test.sh configs/hr32.py work_dirs/hr32/epoch_12.pth 2 --out work_dirs/hr32/test_result.pkl

Postprocess and make submission

python make_submission.py ../submits/submit001.csv --input work_dirs/hr32/test_result.pkl
# python make_submission.py ../submits/submit001.csv --input work_dirs/hr32/test_result.pkl work_dirs/hr48/test_result.pkl

Inference

python inference.py configs/hr32.py ../data/kuzushiji-hrnetv2p_w32-f391e720.pth ../data/test_images/test_012f99f8.jpg ../data/out.jpg