Giter VIP home page Giter VIP logo

yolov4-1's Introduction

Yolo v4 for pytorch , tensorflow eager mode and onnx (by trident api)

thanks for

All the pretrined model weights and cfg is from official site:

https://github.com/AlexeyAB/darknet yolov4.cfg
yolov4.weights

And thanks ultralytics's project, it's really great and helpful.
https://github.com/ultralytics/yolov3

results by yolo v4

Let's look the excellent performance about yolo v4 (pytorch backend)!!

Alt text Alt text Alt text

update (5/3): small item enhance

In yolo v4 have the shortage about missing detection of small itemes. I try to fix the shortage. I found out The best way to fix the issue is modify the objectness in stride=8 Yolo Layer (76*76)

Alt text Alt text

all you need to do is set YoloLayer small_item_enhance=True (only effect 76*76 head)

for module in detector.model.modules():
    if isinstance(module,YoloLayer):
        detector.eval()
        module.small_item_enhance=True

update (5/11): onnx video detection

Yolo v4 onnx

this project is under my another project "trident", a higher order api both in pytorch and tensorflow, and I'll open-source soon.

本專案是基於我目前正在開發的另一個專案trident所開發的,它是一個整合pytorch與tensorflow動態計算圖的高階api,很快我就會將它開源,各位可以先從pip下載安裝。

how to install

trident only support python 3.x
If you want to use pytorch api, you need pytorch 1.2 or higher If you want to use tensorflow api (trident version>=0.5.5), you need tensorflow 2.2.0rc0 or higher (because trident tensorflow api is pure eager mode "without keras, without static graph"") You can install it from pip

pip install tridentx   --upgrade  

after installed trident, you can use following syntax to import it and assign the backed.

import os  
os.environ['TRIDENT_BACKEND'] = 'pytorch'  
import trident as T  
from trident import *  

how to use

  1. pytorch_yolo.py: it is just a basic library derived from trident , to define darknet and yolo basic block.

  2. pytorch_darknet.py: we can construction yolo v4 network and load pretrained weights here.

  3. pytorch_infer_yolo4.py: It's a demo to show how to do object detection by yolo v4 model and how trident api to make things easy.

  4. tf_yolo.py: it is just a basic library derived from trident , to define darknet and yolo basic block.

  5. tf_darknet.py: we can construction yolo v4 network and load pretrained weights here.

  6. tf_infer_yolo4.py: It's a demo to show how to do object detection by yolo v4 model and how trident api to make things easy.

  7. onnx_video_infer.py: it's a demo how to use onnxruntime to infer video near real-time.

You also can download my results in pytorch from google drive:
pytorch pretrained model
pytorch pretrained state_dict
tensorflow pretrained model
tensorflow pretrained state_dict
onnx pretrained model

future work

  1. all the bag of freebies and bag of specials in training context.
  2. use yolov4 for custom datasets

yolov4-1's People

Contributors

allanyiin avatar

Watchers

James Cloos 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.