Giter VIP home page Giter VIP logo

yolov7-coco-text's Introduction

Real Time Text Detection

This repository contains the implementation of the real time text detection project for CS7643. The repository is adapted from the YOLOv7 with additional tools and tuning for training on the COCO-Text V2.0 dataset.

Usage

Training

To train the real time text detection project, you will need the following:

  1. Download the COCO-Text V2.0 dataset from here. You will need both the annotations and the MS COCO 2014 images.
  2. Run the convert-coco-to-yolo.ipynb to convert the COCO-Text V2.0 dataset to YOLO ready format for training.
  3. Install the required packages from the requirements.txt file
  4. In the python environment, we can now train the model. An example training command is: python train.py --workers 8 --device 0 --batch-size 16 --data data/coco-textbox.yaml --img 1280 --cfg cfg/training/yolov7-textbox.yaml --weights "yolov7-tiny.pt" --name yolov7-textbox-tiny-1280-rect --hyp data/hyp.scratch.textbox-tiny-1280-rect.yaml --epochs 50 --adam

Inference

To perform inference, you will need the following:

  1. Install the required packages from the requirements.txt file.
  2. Using the trained model weights or downloading the pretrained model weights, we can now run inference in the python environment. An example inference command is: python detect.py --weights runs/train/yolov7-textbox-tiny-1280/weights/best.pt --conf 0.25 --img-size 1280 --source coco-text/images/test/img18000.jpg --device 0

Summary of Results

Model Image Size Special Arguments [email protected] [email protected]:0.95 FPS (GPU) FPS (CPU)
YOLOv7-Tiny 640 0.5485 0.2791 81.2 ± 0.7 38.3 ± 0.2
YOLOv7-Tiny 640 --rect 0.1558 0.049 86.0 ± 0.3 38.8 ± 0.7
YOLOv7-Tiny 640 –-multi-scale 0.5168 0.2563 79.6 ± 0.9 38.1 ± 0.7
YOLOv7-Tiny 1280 0.6821 0.405 44.6 ± 1.6 11.8 ± 0.1
YOLOv7x 640 0.5951 0.3231 47.4 ± 0.5 5.0 ± 0.1
YOLOv7-e6e 640 0.5883 0.3208 n/a n/a

The YOLOv7-e6e model inference failed to run due to compute limitations.

yolov7-coco-text's People

Contributors

wongkinyiu avatar alexeyab avatar ernestong avatar mkhoshbin72 avatar triple-mu avatar linaom1214 avatar taka-wang avatar spacewalk01 avatar ak391 avatar akashad98 avatar sashaalderson avatar alexeysi avatar raymondben avatar dmlon avatar dhiaeddine-oussayed avatar hran2004 avatar ian321 avatar kadirnar avatar kayce001 avatar kivanctezoren avatar m-gangloff avatar rohanpatankar926 avatar samsamhuns avatar hungtrieu07 avatar greatv avatar jpkoponen avatar ksnzh avatar linghu8812 avatar philipp-schmidt avatar superfast852 avatar

Recommend Projects

  • React photo React

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

  • Vue.js photo Vue.js

    🖖 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. 📊📈🎉

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 ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.