GRENOBLE – CEA-Leti picked “Deep tech for edge artificial intelligence (AI)” as its theme for this year's Leti Innovation Days’ annual conference. The event last week, attended by 800 delegates from all over the world, took place during a blistering heatwave that saw outside temperatures of almost 40 degrees Celsius (104°F).
Inside the Minatec campus, where lectures, workshops and demonstrations were held, attendees were reminded of significant challenges they would face by moving AI to the edge.
Driving such a move is none other than data privacy issues. Other factors prompting the industry to embrace edge AI, however, are their desire to reduce AI computation both in time and energy consumption. Data transfer and memory access would account for up to 90% of system energy usage, thus reducing data movement between memory and processor becomes critical.
CEA-Leti, whose research focus includes health, environment and energy, strutted its stuff by showcasing a number of research projects and demonstrators during the event. Brought to the show floor were Leti's own research projects and others by startups working with or spun out of Leti.
We will share five projects that caught our eyes in the following pages.
Spirit Chip (Photo: EE Times)
Apneaband – wearable device tracking sleep apnea
Spirit – spiking neural networks enabling massively parallel, low power and low latency computation
Bacram – lens-free imaging to quickly identify biological terror threats
Dneuro – high performance DNN running standalone on FPGA
Worms – adding tactile interaction to everyday objects