Translation Systems
The documents describe step-by-step how you can possibly come up with a small working prototype of your own using IBM Watson Services and Python.
Select one of the topics of your interest from the two/three topics below and create a working prototype & submit on Canvas that displays the Classification output for Visual Recognition module and the step-by-step output and also the Code for the Speech to Speech Translation System.
1. Building your own Custom Visual Recognition Classifier: Project1_VR.pdf
2. Building Speech to Speech Translation System: Project2_S2S.pdf
3. If you feel confident then based on this coming weeks’ discussions you can also think of Building a Chatbot for a specific problem. If you decide to work on a chatbot then your submission will be snapshots of each component of the bot with the tested outputs. Here is a link to creating a chatbot on your own local system, instead of on a cloud service. If you choose doing this, you need to customize it to fit your specific problem.
https://towardsdatascience.com/how-to-build-your-own-chatbot-using-deep-learning-bb41f970e281
While there is a github link included to download the files created by the editor, you are expected to create your own, but can use them as a reference.
This is just to get you started so that you feel comfortable linking the cloud services via APIs using Python and also learn how to build applications using some of the most commonly used services. You can enhance and then make it more complex as you feel more comfortable using all the modules.
Update to NOTE: Looks like IBM changed their mind and have made their Visual Recognition service available to the Lite (free) plan again. This means that you can choose the Build your own Custom Visual Recognition Classifier option for the project if you like; without incurring any cost.
NOTE: For the Build your own Custom Visual Recognition Classifier option, there is a problem with it now as IBM Cloud no longer offers Visual Recognition for free. They stopped offering it for free on 16 Oct 2020. You can still choose that option if you like, but you will need to upgrade your account from the free lite version to the standard version, which requires you to enter credit card information. If you still want to use this option, it should cost very much, but there will be a charge from IBM cloud for this service. If you want to do a computer vision prototype from another cloud provider (AWS, Google Cloud, Microsoft Azure, etc.) that’s fine. You can also do a computer vision prototype utilizing OpenVino or OpenCV, utilizing your own system, instead of one that is on the cloud. These last 2 options will be more difficult, but you will learn more if you choose to try one of them.
https://software.intel.com/content/www/us/en/develop/tools/openvino-toolkit.html (Links to an external site.)
https://docs.openvinotoolkit.org/latest/index.html (Links to an external site.)
https://opencv.org/ (Links to an external site.)
Sample Solution
Physical Senses
The sensation of sound occurs when the vibrations from sounds enter our ear and cause little hair like structures, called hair cells, within our inner ear to move back and forth. The hair cells transform this movement into an electrical signal that the brain can use. How well a person can hear largely depend on how intact these hair cells are. Once lost, they don’t grow back, and this is no different for blind people. So blind people can’t physically hear well than others. Yet blind people often outperform sighted people in hearing tasks such as locating the source of sounds. The reason for this emerges when we look beyond the sensory organs, at what is happening with the brain, and how the sensory information is processed by it. In blind people, the visual cortex gets a bit “bored” without visual inputs and starts to “rewire” itself, becoming more responsive to information from the other remaining senses.
