keras-video-classifier | Keras implementation of video classifier | Machine Learning library

by chen0040 Python Version: Current License: MIT

X-Ray Key Features Code Snippets Community Discussions(1)Vulnerabilities Install Support

kandi X-RAY | keras-video-classifier Summary

keras-video-classifier is a Python library typically used in Artificial Intelligence, Machine Learning, Deep Learning, Tensorflow, Keras, Neural Network applications. keras-video-classifier has no bugs, it has no vulnerabilities, it has build file available, it has a Permissive License and it has low support. You can download it from GitHub.

Keras implementation of video classifier

Support

Quality

Security

License

Reuse

Support

keras-video-classifier has a low active ecosystem.

It has 103 star(s) with 52 fork(s). There are 7 watchers for this library.

It had no major release in the last 6 months.

There are 11 open issues and 2 have been closed. On average issues are closed in 0 days. There are 1 open pull requests and 0 closed requests.

It has a neutral sentiment in the developer community.

The latest version of keras-video-classifier is current.

Quality

keras-video-classifier has 0 bugs and 0 code smells.

Security

keras-video-classifier has no vulnerabilities reported, and its dependent libraries have no vulnerabilities reported.

keras-video-classifier code analysis shows 0 unresolved vulnerabilities.

There are 0 security hotspots that need review.

License

keras-video-classifier is licensed under the MIT License. This license is Permissive.

Permissive licenses have the least restrictions, and you can use them in most projects.

Reuse

keras-video-classifier releases are not available. You will need to build from source code and install.

Build file is available. You can build the component from source.

Installation instructions are not available. Examples and code snippets are available.

keras-video-classifier saves you 490 person hours of effort in developing the same functionality from scratch.

It has 1154 lines of code, 64 functions and 24 files.

It has high code complexity. Code complexity directly impacts maintainability of the code.

Top functions reviewed by kandi - BETA

kandi has reviewed keras-video-classifier and discovered the below as its top functions. This is intended to give you an instant insight into keras-video-classifier implemented functionality, and help decide if they suit your requirements.

Fit the VGG16 model
Compile the model
Returns the path to the architecture
Returns the path to the config file
Fit the convolutional model
Scans and generates videos for conv2d
Create a CNN model
Extract videos from Conv2D
Load UCF
Download the ucf file from internet
Plot a history plot and save it
Creates a plot showing the accuracy and validation loss
Scans the input directory and extracts features
Extract features from video
Scans the input directory and extracts images
Extract images from a video
Load the model
Scan the UCF with the given labels
Predict class label
Predict the class of the given video
Loads the model
Scan for conv2d

Get all kandi verified functions for this library.

keras-video-classifier Key Features

No Key Features are available at this moment for keras-video-classifier.

keras-video-classifier Examples and Code Snippets

No Code Snippets are available at this moment for keras-video-classifier.

Community Discussions

Trending Discussions on keras-video-classifier

Keras LSTM uni & bidirectional models converted to Tensorflowjs not producing correct inference

QUESTION

Keras LSTM uni & bidirectional models converted to Tensorflowjs not producing correct inference

Asked 2019-Feb-21 at 13:19

TensorFlow.js version used

tensorflow 1.12.0
tensorflow-base 1.12.0
tensorflow-gpu 1.12.0
tensorflow-hub 0.2.0
tensorflowjs 0.8.0

Browser version used

Firefox 65.0 (64-it) on Windows 10
Microsoft Edge 42.17134.1.0 on Windows 10

Problem Description

I have created & trained a Keras based LSTM bidirectional model in Python to classify video. This model works awesome and classifies the videos with 90+ accuracy. But when I converted this model to tensoflorjs model using the tensorflorjs_converter tool and used the same on browser, the model always throws the same output (top 3 results) for any video input - BasketballDunk; prob. 0.860, BalanceBeam; prob. 0.088, BodyWeightSquats; prob. 0.024

I have checked all the inputs, their shape, etc. that are given to the LSTM bidirectional model and can't find any issues. But still the inference from LSTM bidirectional model is always the same irrespective of the video input. I have ensured that every individual video frame sent to LSTM model as a sequence is correct. (used MobileNet model to recognize each frame and it does correctly and hence concluding that frames sent to LSTM are perfect) Please help me identifying the issue & fix. All the required details are below.

(entire model is based on the examples given in this github repository by Xianshun Chen (chen0040) ->[https://github.com/chen0040/keras-video-classifier])

Details of the model:

uses MobileNet model to extract features
uses LSTM bidirectional model to take-in extracted features and classify the video as one of 20 classes

Dataset used:

UCF101 - Action Recognition Data Set (http://crcv.ucf.edu/data/UCF101.php)

Tensorflowjs converted model:

tensorflowjs converted model, sample videos and html file to test are all in this drive location as zip file: [https://drive.google.com/open?id=1k_4xOPlTdbUJCBPFyT9zmdB3W5lYfuw0]
to test the model, just unzip and build using 'yarn' and run using 'yearn watch'
index.html has the instructions to test

NOTE: I have tried LSTM model (unidirectional) and same issue is with that converted model as well. Only difference is that it produces 'Billards' as the top prediction with probability over 0.95.

Code to reproduce the issue: Code & test artifacts are in a zip file at this Drive location - [https://drive.google.com/open?id=1k_4xOPlTdbUJCBPFyT9zmdB3W5lYfuw0]

...

ANSWER

Answered 2019-Feb-21 at 13:19

Found out the reasons for tfjs converted model not producing the correct inference...at last :)

Reasons:

List item Input to LSTM model had NaN in them! Though I was passing the extracted features from MobileNet model to LSTM, features .dataSync() was not used. Because of this, when I added the extracted features into a tf.buffer they were added as NaN. (when I printed values in log just before adding to tf.buffer, they printed values correctly!...this is strange). So, when I used dataSync() on the extracted features, they got added into tf.buffer correctly.
List item Use of tf.buffer() to store the extracted features (from MobileNet) and converting them to tensors before passing to LSTM model. Instead I used tf.stack() to store the extracted features and then passed the stacked tensor to LSTM model. (I understand that tf.stack() does the equivalent of np.array())

Hope these inputs help someone.

Regards, Jay

Source https://stackoverflow.com/questions/54642294

Community Discussions, Code Snippets contain sources that include Stack Exchange Network

Vulnerabilities

No vulnerabilities reported

Install keras-video-classifier

You can download it from GitHub.
You can use keras-video-classifier like any standard Python library. You will need to make sure that you have a development environment consisting of a Python distribution including header files, a compiler, pip, and git installed. Make sure that your pip, setuptools, and wheel are up to date. When using pip it is generally recommended to install packages in a virtual environment to avoid changes to the system.

Support

For any new features, suggestions and bugs create an issue on GitHub. If you have any questions check and ask questions on community page Stack Overflow .

Find more information at: