Research/Blog
CellStrat > Research/Blog > Artificial Intelligence > Computer Vision > CellStrat AI Lab – Minutes From 23rd. Mar, 2019 Bangalore Meetup
CellStrat AI Lab – Minutes From 23rd. Mar, 2019 Bangalore Meetup
- March 25, 2019
- Posted by: CellStrat Editor
- Category: Artificial Intelligence Computer Vision Deep Learning Machine Learning
No Comments
A fabulous AI Lab session took place on last Saturday in Bangalore. Session saw participation of 35 attendees of which 10 attended via hangout from across the globe. Almost 15 attendees were new and many have shown interest in attending future sessions or joining CellStrat programs. Many non-technical people have started attending and joining the AI bandwagon.
The cumulative expertise of our AI Lab group continues to rise dramatically week after week. We hope to achieve our goal of becoming a world class AI Research Lab in the near future.
- Vivek started with an introductory session on Reinforcement Learning including reward maximization and policy gradient driven learning.
![](http://www.cellstrat.com/wp-content/uploads/2019/03/Vivek_230319_RL-1024x553.jpeg)
- Shreyas (Team Lead for Driverless Cars and RL groups) presented an interesting session on Reinforcement Learning simulators including Mujoco and Carla which are used to train driverless cars. He also discussed Future of AI.
![](http://www.cellstrat.com/wp-content/uploads/2019/03/Shreyas_230319_AI-Future-1024x512.jpeg)
![](http://www.cellstrat.com/wp-content/uploads/2019/03/Shreyas_230319_Carla-1024x512.jpeg)
- Saurabh (Team Lead for Generative Modeling group) presented a rich session on GAN2.0, where one trains specialized GANs by mixing style with noise in various layers of the Generator network. This allows the Generator to control the effect of a particular style — for example high-level facial features such as eyes and hair styles — without changing any other features. It draws inspiration from Style Transfer techniques (style-mixing). GAN2.0 were introduced by Nvidia and can be used for real-like Face Generation.
- Anshumaan (Team Lead for Vision group) presented a deep session on Image Localization and Classification concepts. He discussed differences between traditional RCNN and YOLO – YOLO is a single-algorithm Image Localizer and Classifier as opposed to traditional RCNN/Regional Proposal Network which does the same in multiple algorithmic steps. Anshumaan presented details of YOLOv2 and YOLOv3 which are sophisticated algorithms for this task and also showed code demo on YOLOv3.
![](http://www.cellstrat.com/wp-content/uploads/2019/03/AILabB_230319_Collage.png)