Early warning systems play a key role in multiple business contexts – they could help prevent injuries/deaths at construction sites, predict equipment failure in advance, or determine when a customer is likely to churn. While traditional data analytics approaches to building early warning systems have relied on structured data, new techniques are emerging that use artificial intelligence algorithms on unstructured data – which could be in the form of text, speech, audio, video, etc.

As you participate, you will learn how existing infrastructure can be coupled with AI algorithms to deliver improved safety & performance in different business scenarios.

1 favorite thumb_down thumb_up 2 comments visibility_off  Remove from Watchlist visibility  Add to Watchlist

Outline/Structure of the Talk

1. Introduction to early warning systems

2. Case 1 - Sensor data - detecting rail wheel failure

4. Case 2 - Speech processing - predicting customer churn from conversations, assessment of patient health from conversations.

1. Separating multiple speakers in audio - Diarization - BIC criterion, Phonetics based models

2. Speech to text translation - Wavelet based methods, Pocket Sphinx (open source speech to text tool)

3. Processing the text - Named entity recognition

5. Video analytics - safety monitoring from CCTV footage, automated track inspection.

1. DNNs for image feature extraction - VGG19, ResNet, Inception

2.Object detection techniques - FasterRCNN, YOLO, MaskRCNN

Learning Outcome

1. Learn about AI algorithms and steps involved for object detection & speech processing.

2. Real world applications for safety and performance.

Target Audience

Data Scientists & ML engineers building applications using AI.

schedule Submitted 4 months ago

Public Feedback

comment Suggest improvements to the Speaker
  • Dr. Vikas Agrawal
    By Dr. Vikas Agrawal  ~  3 months ago
    reply Reply

    Dear Sri Vallabha: It will be really nice for us to see details of what specific algorithms you expect to discuss in depth and what results are simply presented in summary. Would you need more time for this talk to cover all the material in-depth as you showed just one of the bullet points as a talk at DataHack? Warm Regards, Vikas

    • Sri Vallabha Deevi
      By Sri Vallabha Deevi  ~  3 months ago
      reply Reply

      Dear Dr. Vikas Agrawal & ODSC 2019 team, I have updated details of algorithms that will be discussed in the talk. Another 15 mins would be helpful to bring in some more depth to the topics. Thank you. Sri Vallabha.