Light

inerplat / joyuriz-classifier Goto Github PK

View Code? Open in Web Editor NEW

9.0 9.0 5.0 17.41 MB

Classify Joyuris using CNN deep learning

License: MIT License

Python 11.13% Jupyter Notebook 88.87%

cnn deep-learning faces image-classification image-processing keras opencv

joyuriz-classifier's Introduction

Hi there 👋

🌱 I’m currently learning Kubernetes, Spring-Kotlin

joyuriz-classifier's People

Contributors

Stargazers

Watchers

Forkers

lindpier samshin7 codingdingdong joshworld hangj11

joyuriz-classifier's Issues

It can't properly handle images of multiple people

In codes:

    x, y, w, h = faces[0]
    crop  = img[y:y+h, x:x+w]

If cv2 lib has find multiple people, It just use one people to process and predict.

The distribution of the number of images is uneven.

The number of train data is uneven, so there is a possibility of bias.
I need to like Yaena more.🐥🥰

Image preprocessing is not appropriate.

In this code, if can't find front faces then try to find profile faces.
So front image is prioritized in a higher than profile image.

    frontFaces = frontFaceCascade.detectMultiScale(gray, 1.3, 5)
    profileFaces = profileFaceCascade.detectMultiScale(gray, 1.3, 5)
    if len(frontFaces) > 0:
        faces = frontFaces
    elif len(profileFaces) > 0:
        faces = profileFaces

It is not appropriate to prioritize face angle of image.

Face pre-processing error

Face detection doesn't work for a re-captured picture of an LCD screen called a "preview picture."

Image file open error

If the file extension differs from the contents of the actual file, an error occurs.
Don't trust the extension when using file.

ex) '.bmp' image file has '.jpg' extension, you can open this file on image viewer(windows default) but your code makes error

CNN doesn't working

acc and loss aren't proceed in the desired direction😨

Epoch 1/50
15/15 [==============================] - 29s 2s/step - loss: 8.6590 - acc: 0.4667 - val_loss: 10.0290 - val_acc: 0.3778
Epoch 2/50
15/15 [==============================] - 22s 1s/step - loss: 10.0290 - acc: 0.3778 - val_loss: 8.5963 - val_acc: 0.4667
Epoch 3/50
15/15 [==============================] - 23s 2s/step - loss: 8.9545 - acc: 0.4444 - val_loss: 8.9545 - val_acc: 0.4444
Epoch 4/50
15/15 [==============================] - 23s 2s/step - loss: 9.3127 - acc: 0.4222 - val_loss: 9.3127 - val_acc: 0.4222
Epoch 5/50
15/15 [==============================] - 26s 2s/step - loss: 9.6709 - acc: 0.4000 - val_loss: 10.3872 - val_acc: 0.3556
Epoch 6/50
15/15 [==============================] - 33s 2s/step - loss: 9.3127 - acc: 0.4222 - val_loss: 11.2452 - val_acc: 0.3023
Epoch 7/50
15/15 [==============================] - 24s 2s/step - loss: 8.9545 - acc: 0.4444 - val_loss: 9.6709 - val_acc: 0.4000
Epoch 8/50
15/15 [==============================] - 23s 2s/step - loss: 10.7454 - acc: 0.3333 - val_loss: 9.6709 - val_acc: 0.4000
Epoch 9/50
15/15 [==============================] - 23s 2s/step - loss: 9.3127 - acc: 0.4222 - val_loss: 8.5963 - val_acc: 0.4667
Epoch 10/50
15/15 [==============================] - 22s 1s/step - loss: 7.8800 - acc: 0.5111 - val_loss: 8.5963 - val_acc: 0.4667
Epoch 11/50
15/15 [==============================] - 22s 1s/step - loss: 10.7454 - acc: 0.3333 - val_loss: 9.3710 - val_acc: 0.4186
Epoch 12/50
15/15 [==============================] - 25s 2s/step - loss: 10.4174 - acc: 0.3537 - val_loss: 9.3127 - val_acc: 0.4222
Epoch 13/50
15/15 [==============================] - 24s 2s/step - loss: 8.5963 - acc: 0.4667 - val_loss: 10.7454 - val_acc: 0.3333
Epoch 14/50
15/15 [==============================] - 23s 2s/step - loss: 7.5218 - acc: 0.5333 - val_loss: 8.9545 - val_acc: 0.4444
Epoch 15/50
15/15 [==============================] - 25s 2s/step - loss: 11.4618 - acc: 0.2889 - val_loss: 9.3127 - val_acc: 0.4222
Epoch 16/50
15/15 [==============================] - 23s 2s/step - loss: 8.9545 - acc: 0.4444 - val_loss: 7.1219 - val_acc: 0.5581
Epoch 17/50
15/15 [==============================] - 23s 2s/step - loss: 10.0290 - acc: 0.3778 - val_loss: 11.4618 - val_acc: 0.2889
Epoch 18/50
15/15 [==============================] - 23s 2s/step - loss: 9.3127 - acc: 0.4222 - val_loss: 9.6709 - val_acc: 0.4000
Epoch 19/50
15/15 [==============================] - 22s 1s/step - loss: 9.6709 - acc: 0.4000 - val_loss: 8.9545 - val_acc: 0.4444
Epoch 20/50
15/15 [==============================] - 23s 2s/step - loss: 11.4618 - acc: 0.2889 - val_loss: 9.3127 - val_acc: 0.4222
Epoch 21/50
15/15 [==============================] - 23s 2s/step - loss: 10.0290 - acc: 0.3778 - val_loss: 9.7458 - val_acc: 0.3953
Epoch 22/50
15/15 [==============================] - 22s 1s/step - loss: 8.2381 - acc: 0.4889 - val_loss: 10.3872 - val_acc: 0.3556
Epoch 23/50
15/15 [==============================] - 24s 2s/step - loss: 9.3127 - acc: 0.4222 - val_loss: 10.3872 - val_acc: 0.3556
Epoch 24/50
15/15 [==============================] - 24s 2s/step - loss: 9.2929 - acc: 0.4234 - val_loss: 8.2381 - val_acc: 0.4889
Epoch 25/50
15/15 [==============================] - 24s 2s/step - loss: 10.0290 - acc: 0.3778 - val_loss: 9.3127 - val_acc: 0.4222

Recommend Projects

React

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

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

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

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
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.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

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

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.