iWildCam_2019_FGVC6
Top 3% (7/336) solution for iWildCam 2019 competition (Categorize animals in the wild), which is as part of the FGVC6 workshop at CVPR 2019
Thanks to my team members!
Requirements
- Python 3.6
- pytorch 1.1.0
About the Code
1. Prepare Data
Download the competition data according to here
After downloading, record the image-file in CSV format.
python prep_data.py
2. Detect and Crop the Image
python detect_crop_image.py
In my method, I first run object detection and crop the bounding box for classification.
3. Train the Model
python train_model.py
4. Prediction
python infer.py
About the Method
I got the best single model predicion result (f1=0.224 in private LB) with following configurations.
model: efficientnet_b0 (imagenet pretrained)
image augmentation: traditional image augmentation + CLAHE + gray scale + cutout + mixup + label smoothing
Please view the detailed report iwildcam_2019_report.pdf