Wearable Technology and Machine Learning for Sustainable Dairy Farming ; The Quest for Open-Source tinyML Heterogeneous Hardware Acceleration: A 10+ Year PULP Journey
Martedi 21 Maggio, Aula Ciminiera (5° piano), DAUIN, si svolgeranno i seguenti seminari:
- ore 10.30 Wearable Technology and Machine Learning for Sustainable Dairy Farming, relatore Prof. Younhyun Kim, Purdue University (Indiana, USA)
Abstract: Heat stress, a consequence of climate warming and the heightened milk production demands on dairy cattle, poses a significant threat to the welfare of the animals and the overall sustainability—economic, environmental, and social—of dairy farming worldwide. Identifying cows experiencing heat stress promptly is essential for enhancing animal welfare, minimizing milk production losses, and conserving water and energy resources required for cooling. This talk introduces our multidisciplinary research on precision dairy farming using wearable technology and machine learning. I will introduce our novel system for detecting heat stress in a timely manner and machine learning approaches on multi-modal sensing. I will talk about our multi-phase development and evaluation of the proposed system and present how we collaborate with animal scientists and biosystems engineers for closed-loop control of the barn.
- ore 11.30 The Quest for Open-Source tinyML Heterogeneous Hardware Acceleration: A 10+ Year PULP Journey, relatore Prof. Francesco Conti, Università di Bologna
Abstract: In the last few years, our perception of what constitutes a "tinyML device" has shifted from simple microcontrollers to complex heterogeneous SoCs suited to execute DNNs directly at the extreme edge in real time and at minimal power cost. These devices provide ultra-low latency and high energy efficiency necessary to meet the constraints of advanced use cases that cannot be satisfied by cloud solutions. However, how can tinyML hardware keep up with the evolution of the AI landscape, continuously pushing towards much larger and more complex models? The costs to develop new accelerators and NPUs for each evolutive step in AI are hard to sustain. A possible way forward is given by the open-source model for digital hardware, popularized by RISC-V: multiple actors - both academic and industrial - collaborate on the development of digital technology that can benefit all parties. In this presentation, I discuss a 10+-year "quest" to push the performance and energy efficiency of tinyML further by exploiting a fully open-source model based on the PULP Platform initiative. I show how the open-source cooperative model makes it possible to combine different ideas and contributions in a technologically portable way, acting as an innovation catalyst and enabling the fast pace of evolution required to keep up with new ideas in AI within a tiny power budget.
(127 KB)