Light

yoosan / i3d-tensorflow Goto Github PK

View Code? Open in Web Editor NEW

49.0 3.0 10.0 32 KB

Inflated 3D ConvNets for video understanding

License: MIT License

Python 100.00%

c3d inception resnet

i3d-tensorflow's Introduction

I3D-TensorFlow

This repo contains the inflated version of the recently popular ConvNets.

Convert weights from the pretrained model

DeepMind have provided the Inception-v1 inflated 3D model, building upon the sonnet. I slightly modified their code and rewrited the i3d model using the protogenetic tensorflow op. The pretrained weights from kinetics-i3d can be easily migrated to the new model. To botain the weights from kinetics-i3d, execute the following instructions.

$ git clone https://github.com/yoosan/i3d-tensorflow
$ cd i3d-tensorflow
$ git clone https://github.com/deepmind/kinetics-i3d
$ python convert_weights.py

Training the I3D model on UCF101

Now I'm preparing the code for training model on the UCF101 dataset. Using the kinetics pretrained weights, we achive a result of 95.2% top-1 accuracy on split 1 of UCF101 with RGB modality.

i3d-tensorflow's People

Contributors

Stargazers

Watchers

Forkers

bityangke willdamon wanjinchang zkk0911 bmyan lfsblack tailin1009 sangilbert chenmoshushi cummins-orgs

i3d-tensorflow's Issues

ModuleNotFoundError: No module named 'snets'

where is the codes for mudule 'snets' ?
from snets.net_utils import unit3D

Compared with P3D

Hello, yoosan,

I am new to do research on video understanding. I want to compare the performance of these two network, both P3D and I3D based on ResNet, and which one is better? Their architectures are similar.

Did you have any understanding about them?

Thx!

Request for fine-tune details~

Hi, yoosan! Recently, I am doing research on I3D, wanting to make the recurrence of this architecture. But I am confused when I am faced with fine-tuning the pre-trained model on UCF-101, whose result only reach 90.8% when input is only RGB.
Can you teach me the details of fine-tuning this model?
Thanks in advance!

fine-tune my own video dataset

hello yoosan:
I plan to classify the 'fight behavior' with fine-tuning the pre-trained I3D model on my own dataset. Can you release the training and test code? thanks a lot!

training r3d on UCF101

Hi, Yoosan, thanks for your code.

Did you train the r3d model on UCF101? If so, what is the accuracy you got?

Can you tell me more detail about training on UCF-101

I train I3d model on UCF-101 split 1 with pretrained Kinetics-imagenet weights, but I only can get the accuracy on test split about 45% , but train accuracy about 95% on rgb modality.
So how did you get the 95.2% accuracy on split 1 , can you give me some advise?

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.