A Con­fig­urable and Ef­fi­cient Mem­ory Hi­er­ar­chy for Neural Net­work Hard­ware Ac­cel­er­a­tor

by Oliver Bause, Paul Palom­ero Bernardo, and Oliver Bring­mann
In MBMV 2024; 27. Work­shop, pages 31–40, 2024.

Con­tri­bu­tion to the De­vel­op­ment of a Smart Video Cap­sule En­doscopy (VCE) Sys­tem for En­hanced Gas­troin­testi­nal Di­ag­no­sis

Bach­e­lor’s The­sis / Mas­ter’s The­sis / Stu­dent Re­search Pro­ject

Ab­stract

Video Cap­sule En­doscopy (VCE) is a di­ag­nos­tic tool for a min­i­mal-in­va­sive ex­am­i­na­tion of the gas­troin­testi­nal tract. As part of an on­go­ing re­search pro­ject, we are de­vel­op­ing a Smart VCE pro­to­type that in­te­grates AI-as­sisted image analy­sis, real-time sen­sor data pro­cess­ing, on-de­vice de­ci­sion mak­ing, and mon­i­tor­ing. To test and ver­ify the pro­to­type, we are si­mul­tan­iously build­ing a hard­ware-in-the-loop (HIL) setup.

Po­ten­tial Top­ics

• Im­ple­men­ta­tion of new hard­ware mod­ules in Sys­temVer­ilog
• Ex­tend the HIL setup
• De­vel­op­ment and com­par­i­sion of de­ci­sion mak­ing ap­proaches
• …

Re­quire­ments

• Sys­temVer­ilog
  • Suc­cess­fully at­tended “ESES” or “MAES” (op­tional)
• C/C++
• Python
• De­vel­op­ment of de­vice dri­ver
  • Suc­cess­fully at­tended “PULPA” (op­tional)

HW-ef­fi­cient Im­ple­men­ta­tion of the Viterbi Al­go­rithm for the Video Cap­sule En­doscopy

Mas­ter’s The­sis

Lever­ag­ing the Pico PIO API for Gen­er­al­ized De­scrip­tions of State Ma­chine-Based In­ter­faces

Mas­ter’s The­sis

HW-De­vel­op­ment of a Min­i­mal ML-based Image Pro­cess­ing Pipeline for Video Cap­sule En­doscopy

Mas­ter’s The­sis