Skip to main content

Test Time Augmentation (TTA) wrapper for computer vision tasks: segmentation,classification, super-resolution, ... etc.

Project description

Edafa

GitHub contributions welcome HitCount
Edafa is a simple wrapper that implements Test Time Augmentations (TTA) on images for computer vision problems like: segmentation, classification, super-resolution, Pansharpening, etc. TTAs guarantees better results in most of the tasks.

Test Time Augmentation (TTA)

Applying different transformations to test images and then average for more robust results.

pipeline

Installation

pip install edafa

Getting started

The easiest way to get up and running is to follow example notebooks for segmentation and classification showing TTA effect on performance.

How to use Edafa

The whole process can be done in 4 steps:

  1. Import Predictor class based on your task category: Segmentation (SegPredictor) or Classification (ClassPredictor)
from edafa import SegPredictor
  1. Inherit Predictor class and implement the main function
    • predict_patches(self,patches) : where your model takes image patches (numpy.ndarray) and return prediction (numpy.ndarray)
class myPredictor(SegPredictor):
    def __init__(self,model,*args,**kwargs):
        super().__init__(*args,**kwargs)
        self.model = model

    def predict_patches(self,patches):
        return self.model.predict(patches)
  1. Create an instance of you class
p = myPredictor(model,patch_size,model_output_channels,conf_file_path)
  1. Call predict_images() to run the prediction process
p.predict_images(images,overlap=0)

Configuration file

Configuration file is a json file containing two pieces of information

  1. Augmentations to apply (augs). Supported augmentations:
    • NO : No augmentation
    • ROT90 : Rotate 90 degrees
    • ROT180 : Rotate 180 degrees
    • ROT270 : Rotate 270 degrees
    • FLIP_UD : Flip upside-down
    • FLIP_LR : Flip left-right
    • BRIGHT : Change image brightness randomly
    • CONTRAST : Change image contrast randomly
    • GAUSSIAN : Add random gaussian noise
    • GAMMA : Perform gamma correction with random gamma
  2. Combination of the results (mean). Supported mean types:
    • ARITH : Arithmetic mean
    • GEO : Geometric mean
  3. Number of bits image (default is 8-bits) (bits).

Example of a conf file in json format

{
"augs":["NO",
"FLIP_UD",
"FLIP_LR"],
"mean":"ARITH",
"bits":8
}

Example of a conf file in yaml format

augs: [NO,FLIP_UD,FLIP_LR]
mean: ARITH
bits: 8

You can either pass file path (json or yaml) or the actual json text to conf parameter.

Contribution

All contributions are welcomed. Please make sure that all tests passed before pull request. To run tests

nosetests

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

edafa-0.1.5.tar.gz (7.0 kB view details)

Uploaded Source

Built Distribution

edafa-0.1.5-py3-none-any.whl (9.4 kB view details)

Uploaded Python 3

File details

Details for the file edafa-0.1.5.tar.gz.

File metadata

  • Download URL: edafa-0.1.5.tar.gz
  • Upload date:
  • Size: 7.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.1.1 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.6.0 requests-toolbelt/0.9.1 tqdm/4.35.0 CPython/3.7.3

File hashes

Hashes for edafa-0.1.5.tar.gz
Algorithm Hash digest
SHA256 f8153c2dd83cc31e2175ec95a0806479922065635d36d1eb4cfcad9b730c50df
MD5 24a7434cd32ffd1d82ede8cd4f07799c
BLAKE2b-256 7ae0345050da2e2d8ba9300276c3b91b598284f31816c0bf559752bfd3c6e532

See more details on using hashes here.

File details

Details for the file edafa-0.1.5-py3-none-any.whl.

File metadata

  • Download URL: edafa-0.1.5-py3-none-any.whl
  • Upload date:
  • Size: 9.4 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.1.1 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.6.0 requests-toolbelt/0.9.1 tqdm/4.35.0 CPython/3.7.3

File hashes

Hashes for edafa-0.1.5-py3-none-any.whl
Algorithm Hash digest
SHA256 b716c717a4f07d53c46fa215e7aaf2c84cd08c24ca2ee1998ada31b59eb5f281
MD5 c74dcec90706257d2bc2c59608847add
BLAKE2b-256 44a1c3974524ad587c1cb39a4607a903fc52e2fea846fcef89f2e5123a300b8d

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page