GithubHelp home page GithubHelp logo

qingliwang / gdxray-retinanet Goto Github PK

View Code? Open in Web Editor NEW

This project forked from aurotripathy/gdxray-retinanet

0.0 2.0 0.0 220.99 MB

This project applies the RetinaNet object detector on the GDXray dataset, Castings group.

Python 100.00%

gdxray-retinanet's Introduction

GDXray-retinanet

This project applies the RetinaNet object detector on the GDXray dataset, Castings group.

Reuse

We entirely reuse the Keras Retinanet Object Detection Framework and just apply the dataset as instructed by the framework.

Configuration

We use the frameworks's default Focal Loss hyper-paramters of alpha=0.25, and gamma=2.0

Training

./train.py --multi-gpu=3 --batch-size=3 --freeze-backbone \
--no-evaluation --steps=10000 --epochs=20 --snapshot-path xray3-snapshots csv ../\
utils/annotations-with-negatives/train_annotations.csv ../utils/annotations/classes.csv

Resume training from a snapshot (--snapshot)

./train.py --gpu=1 --freeze-backbone --no-evaluation --steps=10000 --epochs=5 \
--snapshot  xray3-snapshots/resnet50_csv_20.h5 \
--snapshot-path xray4-snapshots \
csv ../utils/annotations-with-negatives/train_annotations.csv ../utils/annotations/classes.csv

Evaluate

./evaluate.py --save-path results/ --max-detections=7 \
csv ../utils/annotations-with-negatives/test_annotations.csv \
../utils/annotations-with-negatives/classes.csv xray-snapshots/resnet50_csv_11.h5

Initial Results

With the default Focal Loss hyperparamter settings, the mAP is 0.76

A sampling of inital results show below.

Ground-truth bounding-boxes are in green, detections are in blue.

Logo Title Text 1

Logo Title Text 1

Logo Title Text 1

gdxray-retinanet's People

Contributors

aurotripathy avatar

Watchers

James Cloos avatar  avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.