ODAch-Object-Detection-ttA in Pytorch

ODA is a test-time-augmentation tool for 2d object detectors.

Used in Kaggle object detection competitions!

Install

pip install odach

Usage

See Example.ipynb.

The setup is very simple, similar to ttach.

import odach as oda
# Declare single scale TTA variations
mono = [oda.VerticalFlip(),oda.HorizontalFlip(), oda.Rotate90(), oda.Multiply(0.9), oda.Multiply(1.1)]
# Declare multiscale-TTA with 0.8~1.2x image sizes.
multi = [oda.MultiScale(i) for i in [0.8, 0.9, 1.1, 1.2]] + [oda.MultiScaleFlip(i) for i in [0.8, 0.9, 1.1, 1.2]]

# load image
impath = "imgs/cars3.jpg"
img = loadimg(impath)
# wrap model and tta
tta_model = oda.TTAWrapper(model, mono, multi)
# Execute TTA!
boxes, scores, labels = tta_model.inference(img)

• The image size should be square.

model output wrapping

• Wrap your detection model so that the output is similar to torchvision frcnn format: [["box":[[x,y,x2,y2], [], ..], "labels": [0,1,..], "scores": [1.0, 0.8, ..]]

Thanks

nms, wbf are from https://kaggle.com/zfturbo

tta is based on https://github.com/qubvel/ttach, https://github.com/andrewekhalel/edafa/tree/master/edafa and https://www.kaggle.com/shonenkov/wbf-over-tta-single-model-efficientdet