Tensorflow is Google’s open source Deep learning Library. You can build a lot of machine learning based applications using this framework along with Python programming language. Being open source, many people build applications or other frameworks over Tensorflow and publish them on Github. In this article, I will share some amazing Tensorflow Github projects that you can use directly in your application or make it better to suit your needs.
Table of Content
- Tensorflow: An intro and getting started
- Neural Style
- Deep Speech
- Sentence Classification
- Image classification
Intro to Tensorflow
If you already know what Tensorflow is, and how it works, I will suggest you skip to the next section. But if you are interested in learning Tensorflow with Python, then I will recommend you to visit this course Python for Data Science and Machine Learning Bootcamp. If you are a beginner, then this is perfect for you. As I mentioned earlier, Tensorflow is a deep learning library. Build and test deep neural networks with this framework. Deep learning has enabled us to build
Deep learning has enabled us to build complex applications with great accuracies. Whether it is to do with images, videos, text, audio, deep learning can solve problems in that domain. Tensorflow can be used to achieve all of these applications. Understanding the Tensorflow Concepts and basics isn’t difficult. The reason for its popularity is the ease with which developers can build, test and deploy machine learning application with tensorflow. The Tensorflow GitHub projects shared in this article are what I have come across to be awesome. As and when I find more, I will keep updating this list.
Project 1: Neural Style
This has to be one of the coolest Tensorflow Github Projects. The neural style is a process of transferring the style of one picture on to another with out losing out on the characteristics of the first picture. In simple terms, with this project, you can have your own Prisma like application built using Tensorflow. The painting style is combined with the lion’s image to get the first image above. You can combine multiple styles onto one image and also decide the percentage of style to be applied.
Tensorflow Github project link: Neural Style TF ( image source from this Github repository)
Project 2: Mozilla Deep Speech
This Tensorflow Github project uses tensorflow to convert speech to text. Speech to text is a booming field right now in machine learning. Applications of it include virtual assistants ( like Siri, Cortana, etc) in smart devices like mobile phones, tablets, and even PCs. People’s accents vary across the world and due to that, speech to text conversions are a difficult topic to crack. However, we can achieve great accuracies with deep learning in place.
Github Link: Mozilla Deep Speech
Project 3: Sentence Classification with Tensorflow
Sentence classification refers to the process of identifying the category of a sentence. For example, if you have a sentence ” The food was extremely bad”, you might want to classify this into either a positive sentence or a negative sentence. This is called sentiment analysis. Now, even in this concept, there are a lot of complexities where categorization of sentences becomes difficult because of the sentence structure. With the convolutional neural networks, we can try to build a strong text based classifier.
Github Link: Sentence classification with CNN
Project 4: Image classification/ Object Recognition
Image classification refers to training our systems to identify objects like a cat, dog, etc, or scenes like driveway, beach, skyline, etc. The scope of computer vision is huge. From face recognition to emotion recognition, to even visual gas leak detection comes under this category. Though the procedures and pipelines vary, the underlying system remains the same.
I have created the following Tensorflow GitHub repository which has two parts associated with it. First part is where you can setup a tensorflow based classifier just to test it out. In the second part, you can train your own models to identify those classes.
GitHub project link: TF Image Classifier with python
Conclusion on Tensorflow Github Projects
I hope that you have found these projects to be awesome. You can build on top of these or use it as it is. The best way to learn is to actually do something. If you are a beginner, then it’s an amazing investment to buy a course and make use of it completely. Here is the link to the course: Python for Data Science and Machine Learning Bootcamp .You can use a few coupon codes “JULY15UDEMY” or “GROUPONUDEMY” or “UJ1L202” which you can use to get up to 95% off. This reduces the price to just $10.
Do you have any Tensorflow Github project? Let me know in the comments below and you might get your project featured on the next update of this blog post!