-
On the Query Strategies for Efficient Online Active Distillation
Authors:
Michele Boldo,
Enrico Martini,
Mirco De Marchi,
Stefano Aldegheri,
Nicola Bombieri
Abstract:
Deep Learning (DL) requires lots of time and data, resulting in high computational demands. Recently, researchers employ Active Learning (AL) and online distillation to enhance training efficiency and real-time model adaptation. This paper evaluates a set of query strategies to achieve the best training results. It focuses on Human Pose Estimation (HPE) applications, assessing the impact of select…
▽ More
Deep Learning (DL) requires lots of time and data, resulting in high computational demands. Recently, researchers employ Active Learning (AL) and online distillation to enhance training efficiency and real-time model adaptation. This paper evaluates a set of query strategies to achieve the best training results. It focuses on Human Pose Estimation (HPE) applications, assessing the impact of selected frames during training using two approaches: a classical offline method and a online evaluation through a continual learning approach employing knowledge distillation, on a popular state-of-the-art HPE dataset. The paper demonstrates the possibility of enabling training at the edge lightweight models, adapting them effectively to new contexts in real-time.
△ Less
Submitted 4 September, 2023;
originally announced September 2023.
-
Speck: A Smart event-based Vision Sensor with a low latency 327K Neuron Convolutional Neuronal Network Processing Pipeline
Authors:
Ole Richter,
Yannan Xing,
Michele De Marchi,
Carsten Nielsen,
Merkourios Katsimpris,
Roberto Cattaneo,
Yudi Ren,
Yalun Hu,
Qian Liu,
Sadique Sheik,
Tugba Demirci,
Ning Qiao
Abstract:
Edge computing solutions that enable the extraction of high-level information from a variety of sensors is in increasingly high demand. This is due to the increasing number of smart devices that require sensory processing for their application on the edge. To tackle this problem, we present a smart vision sensor System on Chip (SoC), featuring an event-based camera and a low-power asynchronous spi…
▽ More
Edge computing solutions that enable the extraction of high-level information from a variety of sensors is in increasingly high demand. This is due to the increasing number of smart devices that require sensory processing for their application on the edge. To tackle this problem, we present a smart vision sensor System on Chip (SoC), featuring an event-based camera and a low-power asynchronous spiking Convolutional Neural Network (sCNN) computing architecture embedded on a single chip. By combining both sensor and processing on a single die, we can lower unit production costs significantly. Moreover, the simple end-to-end nature of the SoC facilitates small stand-alone applications as well as functioning as an edge node in larger systems. The event-driven nature of the vision sensor delivers high-speed signals in a sparse data stream. This is reflected in the processing pipeline, which focuses on optimising highly sparse computation and minimising latency for 9 sCNN layers to 3.36μs for an incoming event. Overall, this results in an extremely low-latency visual processing pipeline deployed on a small form factor with a low energy budget and sensor cost. We present the asynchronous architecture, the individual blocks, and the sCNN processing principle and benchmark against other sCNN capable processors.
△ Less
Submitted 27 May, 2024; v1 submitted 13 April, 2023;
originally announced April 2023.
-
Hyperspectral and LiDAR data for the prediction via machine learning of tree species, volume and biomass: a possible contribution for updating forest management plans
Authors:
Daniele Michelini,
Michele Dalponte,
Angelo Carriero,
Erico Kutchart,
Salvatore Eugenio Pappalardo,
Massimo De Marchi,
Francesco Pirotti
Abstract:
This work intends to lay the foundations for identifying the prevailing forest types and the delineation of forest units within private forest inventories in the Autonomous Province of Trento (PAT), using currently available remote sensing solutions. In particular, data from LiDAR and hyperspectral surveys of 2014 made available by PAT were acquired and processed. Such studies are very important i…
▽ More
This work intends to lay the foundations for identifying the prevailing forest types and the delineation of forest units within private forest inventories in the Autonomous Province of Trento (PAT), using currently available remote sensing solutions. In particular, data from LiDAR and hyperspectral surveys of 2014 made available by PAT were acquired and processed. Such studies are very important in the context of forest management scenarios. The method includes defining tree species ground-truth by outlining single tree crowns with polygons and labeling them. Successively two supervised machine learning classifiers, K-Nearest Neighborhood and Support Vector Machine (SVM) were used. The results show that, by setting specific hyperparameters, the SVM methodology gave the best results in classification of tree species. Biomass was estimated using canopy parameters and the Jucker equation for the above ground biomass (AGB) and that of Scrinzi for the tariff volume. Predicted values were compared with 11 field plots of fixed radius where volume and biomass were field-estimated in 2017. Results show significant coefficients of correlation: 0.94 for stem volume and 0.90 for total aboveground tree biomass.
△ Less
Submitted 30 September, 2022;
originally announced September 2022.
