Policy Indications: UK Graduate route to open to international students on 1 July 2021  |  Leadership Instincts: VP appeals to students to connect their knowledge with social relevance  |  Leadership Instincts: Catherine Dulac receives Nomis Distinguished Scientist and Scholar Award  |  Leadership Instincts: Online school reviews reflect school demographics more than effectiveness  |  Leadership Instincts: Researchers virtually open and read sealed historic letters  |  Cover Story: At Vantage Point  |  Management lessons: Why Aluminium Cans are Great for Packaging of Beverages?  |  Parent Interventions: Motivation to Perform  |  Parent Interventions: Poor Quality Carbs Harmful for Heart  |  Parent Interventions: Beat Covid stress with Yoga  |  Education Information: Suggestions invited on Draft UGC Regulations, 2021  |  Life Inspirations: Little Acts of Kindness  |  Parent Interventions: Travel Local, Play Safe!  |  Art & Literature: IIT Hyderabad to hold techno-cultural fest Elan & nVision 2021  |  Teacher Insights: Minister releases Study Material of Indian Knowledge Tradition Courses  |  
February 03, 2020 Monday 03:19:18 PM IST

Intel AI to Count Antarctic Penguins

Science Innovations

Gramener, an Intel AI Builder member and data science company has come up with a new crowd counting solution to study penguin populations in the Antarctic. Penguins face the serious threats on unprecedented breeding issues for the past three years. This coupled with climate change issues may result in the disappearance of penguins by the year 2100. The Gramener solution uses images from over 40 locations and uses a density-based counting approach to approximate the number of penguins in clusters of different sizes fro the images.

In partnership with Microsoft AI for Earth, Gramener researchers trained a deep learning model to count the penguins. This solution has been repurposed and benchmarked on Intel Xeon Scalable processors and the Intel Optimization for PyTorch for optimized performance. This solution could help researchers overcome challenges in manually counting penguins from camera traps, which can be tricky due to perspective distortion, penguins standing too close together or clustering, and diversity of camera angles.


Comments