Light

elvishelvis / 68-retinaface-pytorch-version Goto Github PK

View Code? Open in Web Editor NEW

116.0 9.0 27.0 31.84 MB

Working on 68/96 landmarks detection with RetinaFace with MobileNet 0.25

Python 100.00%

68-retinaface-pytorch-version's Introduction

Retinaface-Pytorch-version

It's not the best version of my model due to confidentiality

Thanks to Alvin Yang (https://github.com/supernotman/RetinaFace_Pytorch)

This is the branch for 68 landmarks detection, the pre-trained model is in ./out

Working on 96 landmarks detection( refer to the other branch)

The model also predicted the occulded part of the landmarks, can hide them if don't want them to show up.

Based on RetinaFace

current model

mobileNet V1+FPN+context module+ regressor 1.6MB CPU～10FPS GPU 50FPU

Train：（ Please refer to dataloader.py to change the file location）

python3 train.py -train This model use LS3D-W dataset，or change your dataset to the format of demo.pt/ demo.jpg（68*2 tensor）

Use local camera ：

python3 video_detect.py （ need to delete all 'cuda()', and run locally with a CPU）

Eval Model：

python3 train.py -train False

Todo:

Use SBR and BFLD to improve performance

If you have train a model with this code, welcome to discuss with me at [email protected]

68-retinaface-pytorch-version's People

Contributors

Stargazers

Watchers

68-retinaface-pytorch-version's Issues

question about lip weights in loss function

hi.
Thanks for sharing such a nice code.
When I tried your code, I found in losses.py, there are two lines to add lip weighted when compute landmarks loss, but seems idx 99 and idx 37 not the lips idx.

May I ask which two points 99 and 37 represent?

Why my loss is lower?

I use LS3D-W dataset , when I training, the loss is lower .. expamle: total_loss : 1.1131469011306763 classification: 0.70500928 bbox: 0.07543718814849854 landmark: 0.6654008626937866 .
Check whether the fault is rectified.

The landmark loss reduces slowly.

Hello, ElvishElvis.
Thank you for your great work!
I change the backbone to mobilenet_v3 and regnet, and use the LS3D-W dataset as my training dataset. But the landmark branch loss reduces slowly (using smoothL1 Loss), for example, 1.79 -> 1.77 after 20 epochs, while the other losses reduces normally.
So, have you met this problem? And could you share your training strategy if possible?

Hi，你的widerface是重新标注了68点吗

hi，what's the size of input image ，thank you for replaying me

i want to know the input size ，now i have a task that very urgent

So many BUGs in this code

Did you train with wing loss? I see that you wrote this loss function, but it was not used, why?

Training Data

Hello, thank you for your sharing.
I'm developing the retinaface too, could you please provide the training data?
I want to know the content in path1 = "/versa/elvishelvis/RetinaFace_Pytorch/CelebA/Anno/list_bbox_celeba.txt" and path2 = "/versa/elvishelvis/RetinaFace_Pytorch/CelebA/Anno/list_landmarks_celeba.txt"

RuntimeError: Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor) should be the same

RuntimeError: Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor) should be the same

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.