Supervision : Bertrand GRANADO

Co-supervision : HACHICHA Khalil

ZIP-CNN

Digital systems used for the Internet of Things (IoT) and Embedded Systems have seen an increasing use in recent decades. Embedded systems based on Microcontroller Unit (MCU) solve various problems by collecting a lot of data. Today, about 250 billion MCUs are in use. Projections in the coming years point to very strong growth.
Artificial intelligence has seen a resurgence of interest in 2012. The use of Convolutional Neural Networks (CNN) has helped to solve many problems in computer vision or natural language processing. The implementation of CNN within embedded systems would greatly improve the exploitation of the collected data. However, the inference cost of a CNN makes their implementation within embedded systems challenging.
This thesis focuses on exploring the solution space, in order to assist the implementation of CNN within embedded systems based on microcontrollers. For this purpose, the ZIP-CNN methodology is defined. It takes into account the embedded system and the CNN to be implemented. It provides an embedded designer with information regarding the impact of the CNN inference on the system. A designer can explore the impact of design choices, with the objective of respecting the constraints of the targeted application. A model is defined to quantitatively provide an estimation of the latency, the energy consumption and the memory space required to infer a CNN within an embedded target, whatever the topology of the CNN is. This model takes into account algorithmic reductions such as knowledge distillation, pruning or quantization. The implementation of state-of-the-art CNN within MCU verified the accuracy of the different estimations through a measurement process.

2019-2024 Publications

• 2024
  • P. Dobias, Th. Garbay, B. Granado, Kh. HACHICHA, A. Pinna : “Algorithms for scheduling CNNs on multicore MCUs at the neuron and layer levels”, Microprocessors and Microsystems: Embedded Hardware Design, vol. 111, pp. 105107, (Elsevier) (2024)
• 2023
  • Th. Garbay : “ZIP-CNN”, thesis, phd defence 05/13/2023, supervision Granado, Bertrand, co-supervision : Hachicha, Khalil (2023)
• 2022
  • Th. Garbay, Kh. HACHICHA, P. Dobias, W. Dron, P. Lusich, I. Khalis, A. Pinna, B. Granado : “Accurate Estimation of the CNN Inference Cost within Microcontrollers”, TinyML EMEA 2022, Limassol, Cyprus (2022)
  • Th. Garbay, Kh. HACHICHA, P. Dobias, W. Dron, P. Lusich, I. Khalis, A. Pinna, B. Granado : “Accurate Estimation of the CNN Inference Cost for TinyML Devices”, 35^th IEEE INTERNATIONAL SYSTEM-ON-CHIP CONFERENCE, Belfast, Ireland (2022)
  • P. Dobias, Th. Garbay, B. Granado, Kh. HACHICHA, A. Pinna : “Comparative Study of Scheduling a Convolutional Neural Network on Multicore MCU”, DASIP 2022 - 15^th International Workshop on Design and Architecture for Signal and Image Processing, vol. 13425, Lecture Notes in Computer Science, Budapest, Hungary, pp. 69-80, (Springer International Publishing) (2022)
  • Th. Garbay, P. Dobias, W. Dron, P. Lusich, I. Khalis, A. Pinna, Kh. HACHICHA, B. Granado : “Estimation du coût d’inférence des réseaux de neurones convolutifs dans un microcontrôleur”, 16^e Colloque du GDR SoC2, Strasbourg, France (2022)
• 2021
  • Th. Garbay, P. Dobias, W. Dron, P. Lusich, I. Khalis, A. Pinna, Kh. HACHICHA, B. Granado : “CNN Inference Costs Estimation on Microcontrollers: the EST Primitive-based Model”, 28^th IEEE International Conference on Electronics Circuits and Systems IEEE ICECS 2021, Dubaï, United Arab Emirates (2021)
• 2019
  • Th. Garbay, O. Chuquimia, A. Pinna, H. Sahbi, X. Dray, B. Granado : “Distilling the knowledge in CNN for WCE screening tool”, 2019 Conference on Design and Architectures for Signal and Image Processing (DASIP), Montreal, Canada, pp. 19-22, (IEEE) (2019)
  • O. Chuquimia, Th. Garbay, W. Xu, A. Pinna, X. Dray, H. Sahbi, B. Granado : “Study to integrate CNN inside a WCE to realize a screening tool”, Journées d'Etude sur la TéléSanté, Paris, France (2019)