READ ME File For 'TinyOps: ImageNet Scale Deep Learning on Microcontrollers'

ReadMe Author: SULAIMAN SADIQ, University of Southampton

This dataset supports the publication:
AUTHORS: Sulaiman Sadiq, Jonathon Hare, Partha Maji, Simon Craske, Geoff Merrett
TITLE: TinyOps: ImageNet Scale Deep Learning on Microcontrollers
WORKSHOP: CVPR 2022 Efficient Deep Learning for Computer Vision
DOI: https://doi.org/10.5258/SOTON/D2188

This dataset contains:

---- Fig-3-Top ----

Data to make bar chart top in Figure 3. The figure shows how overlaying different types of frequently accessed data reduces latecny of TinyOps compared to external memory

Filename: 'Fig3a.csv'

Data: ID of data that is overlayed from SDRAM to SRAM. Latency reduced by overlaying data specified in ID from SDRAM to SRAM


---- Fig-3-Bot ----

Data to make bar chart bottom in Figure 3. The figure shows much SRAM is consumed by overlaying different types of frequently accessed data

Filename: 'Fig3b.csv'

Data: ID of data that is overlayed from SDRAM to SRAM. SRAM usage of overlaying data specified in ID from SDRAM to SRAM

---- Table-1 ----

Internal and external memory speicifications of platforms used in experiments in section 3 of the paper

Filename: 'Table1.csv'

Data: Name of the platforms followed by the memory specifications, specifically the amount of internal sram and flash in addition to external sdram and flash

---- Table-2 ----

Comparison of performance of models from the internal memory design space and tinyops design space. measured stats include models stats and the deploment stats on platforms given in Table 2

Filename: 'Table2.csv'

Data: Name of the platforms followed by the stats of the models deployed on the platforms including name, design space, size, ram, MACs. for each of the models we meausre int8 accuracy, latency, current and energy per inference when deploying on the given platforms

---- Table-3 ----

Comparison of energy consumption of the same model when running on different platforms under different memory configurations

Filename: 'Table3.csv'

Data: Name of the platforms followed by the performance stats measure by deploying on the platforms including latency, current consumption and energy per inference. 

Date of data collection: 2022

Licence: CC BY

Related projects:

This work was supported by the UK Research and Innovation (UKRI) Centre for Doctoral Training in
Machine Intelligence for Nano-electronic Devices and Systems [EP/S024298/1] and the Engineering
and Physical Sciences Research Council (EPSRC) International Centre for Spatial Computational
Learning [EP/S030069/1]. The authors also acknowledge the use of the IRIDIS High Performance Computing
Facility, and associated support services at the University of Southampton, in the completion of this
work. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising.

Date that the file was created: Apr, 2022