Global-Wheat-Detection

kaggle - https://www.kaggle.com/c/global-wheat-detection

Description

• ssd model
• random rotate90 and other augmentation for train
• wandb logger, used pytorch-lightning/pytorch, fp16

In the making(missed to do):

• training on 5 folds(clustered by kmeans, previously reduced by PCA)
• tta for test(inference)

Test results

in /test_output/viz

Submission score [email protected]:0.05

score: 0.6448

kaggle notebook - https://www.kaggle.com/heorgiibolotov/kernel1a17c2da7e

cut version of logs while training

Project structure

├── apex.sh              
├── eda.ipynb   
├── image_plot.py  
├── prepare_data.py  # invalid version 
├── requirements.txt  
├── submission.csv  
├── wandb
├── cp_data_from_hdd.sh  
├── engdata.py # invalid version
├── parse_data.py # invalid version
├── report         

├── retinaface    
  ├── box_utils.py  
  ├── data_augment.py
  ├── inference.py
  ├── make_submit.py
  ├── net.py
  ├── prior_box.py  
  ├── train.py
  ├── configs       
  ├── dataset.py       
  ├── __init__.py 
  ├── multibox_loss.py
  ├── network.py 
  ├── __pycache__   
  ├── utils.py
  
├── test_output
  ├── labels
  ├── viz
  
├── data
  ├── annotations            
  ├── test   
  ├── train.csv 
  ├── train_with_bbox
  ├── sample_submission.csv
  ├── train  
  ├── train.json
  ├── train_without_bbox

Usage

# step 1.(in this repo invalid version of the code)
!python engdata.py  

# step 2. 
!python -m retinaface.train -c retinaface/configs/2020-07-20.yaml

# step 3. 
!python -m retinaface.inference -i data/test -c retinaface/configs/2020-07-20.yaml -o test_output -w 2020-07-27/epoch=7.ckpt -v