ProCam Low-Cost Open Thermal Temperature Project

By Charles Rollet, Published May 12, 2020, 08:32am EDT

An engineering professor in Switzerland is building what he hopes will be the world's lowest-cost fever camera, costing a fraction of commercial offerings today.

IPVM Image

In this post, based on an interview the project's founder Touradj Ebrahimi, we examine the:

  • Professor's Background
  • ProCam Timeline, Business Model Details
  • ProCam Purpose: "Very Low Cost", Several Hundred Dollars
  • Multiple Cheap Sensors Key to Low Cost
  • ProCam AI Use Details
  • Number of Sensors Currently Unknown
  • Accuracy Unknown So Far
  • ProCam Won't Measure Specific Increments
  • Effective Distance Of 7 Meters
  • Long List of Challenges

******, ***** *** *** been ***** ****** ***, is ***** ** ******* this ** ********* ******* low-quality ******* *** ***** AI ** ****** *** "the *****".

**** ** ********* ** any ***-**** ***** *** solution, ** ***** ********* is ******* ******* ** the **** *********** ******* 'normal' *** '*****' **** the ******* ********** ** measuring ****** ** *** wild.

*******, ******'* ******* **** it's ************* ******** ** create **** * ******, and *********** **** *** accuracy **** ** *****, says *** ****** **** only **** ******* ******* has ** "*****-*******" *********** or *** (** * certain ****** ** **********).

Background ** *********

IPVM Image

******* ********** ** ******* ********* at *** ***** ******* Institute ** **********, ******** (EPFL), *** ** ********'* *** *********** *******.********, *** *** * Ph.D. ** ********** ***********, has ************ *********** ** video *********** ****** **** thermal *******. ** ** also *** **-******* ** startup*********, ***** *********** ** using ******** ******* ** once ** ****** ** videos *** *************.

ProCam ********, ******** *****

****** ** **** *** - ** *** ******* in ***** ** **** of ** ****-*********** ********* and *** * ************** ** ******** *** his ********.

****** **** **** * GitHub ********** ******** **********:

*** ***** ********** *** scheduled *** **** *** the ***** **** *** "October, ********, ** ********", Ebrahimi ****. ** ***** testing *** *** **** performed *** ******** ******* to ** ** ** the ****** ******.

****** ** * "*******, not * *******" - it ** ***-***-****** *** there ** ** **** to **** ***********/**** *******. Rather, ** *** ****** proves **********, *** **** is ** ******* *** software/hardware ******* *** **** to ***** *** ****, although ** ***** ****** any ************ ******* ***** scale *****************.

ProCam *******: "**** *** ****", ******* ******* *******

******** **** **** *** ProCam ***** ** ****** at "******* ******* *******" so **** ****** ****** could *** **. *** aim ** ********** *** in ***** *****, *******, and ***** ****** **** can't ****** ****-*** ******* cams:

** **** ** *** if **'* ******** ** get * **** *** cost [*****] ******. **** of ***** ******* *** very, **** *********.

*** ***** **** ******* of ****** ** *******, which ******** **** **** be ******** ** ********** footage ** * ******* server *** **** ******** the ******** ****** ** health *********** ** **-********* relevant ********.

Multiple ***** ******* *** ** *** ****

** *** *****, ****** uses ******** ***** ******* and ***** ****** *********, off-the-shelf **********:

*** ****** ** *** using ***** *** ** one **** ** **** popular, *** *** *****, which ** ** * 24 ******. ** ***** Francs, **'* ***** ** CHF *** ****** [$**]. And *** **** ** the ****** ** ********* Pi, ********* * ***** PC.

****** ******** ** ******** ** ±2°C, ***** ** ********* far *** *****- **** fever *** ************* **** accuracy ** ±*.*°* ******* Celsius, *** ************* ********* specify * ******* ±*.*°* of ********.

IPVM Image

ProCam ** *** *******

******'* ******* ** ** use ** ** ************* improve **** **** ******** level:

** *** *** ***** the ***** **** ** generated **** **** ** the *******, *** *** actually **** **** *** with **** ** ********* and ******* **********, *** can ****** **. ** your ****** *** ****/***** 3 [*******] *******, *** will ** **** ** go **** ***** ** 0.5 ** ***** ****. It ******* ** *** you ***** *** ***** and *** **** ******* you ****.

****** **** *** * superresolution ********* ** ******** resolution *** ******* ** filter *** *** *****:

** *** ** **-***** superresolution ******** ***** ** also **** ****** **** can ** ** ** 8 ***** ********** *** the ******* **** ** be ***** **** - you *** *** ** 8 * *, ** every ****** ***** ******* 64 ****** [...] ** fuse *** **** **** each ****** ** ***** to ******* *** ******** of **** ****** ** modelling *** ***** ******* to *** ******* **** as ***** ***** ** Wigner-Ville ************ **”

**** ***** ******** ******* he **** ** ********** studies ** ********** ******** showing **** ** *** be **** ** **** way ************ *** ***** cameras, ******* ** ***** not **** ***.

******: ******** ******* *** that********* *** ********* * system*** * ****** ******** that ************** ******* ******* someone *** * **** and ***** ***** ***********. However, ***** ** ** evidence **** ********* ****** uses ** ** ******** accuracy, ***** ** **** ProCam ********.

Number ** ******* ********* *******

****** ********* *** * sensors *** ******* ** hasn't **** ****** ***, Ebrahimi *****'* **** *** many ** **** ******** end ** ******:

***** *** **'* ***** low-cost *******, ****'* *** number ** *** ****** but ***** *** ** enough ** ***** *** need ****. ** ***'* know, ** **** ***. One ***** **** ** known, *** * *** of ****** ***** ********* have **********, ** **** the **** ******* ** the **** ****, *** better *** ******* *** accuracy.

Accuracy ******* ** ***

******** **** *** *** know *** ******** ****** will ** ** ** has *** **** ****** yet:

* ***'* **** **** will ** *** ********, that's *** ****, *** I ***'* **** ******* false **** ***** ** do *****.

ProCam ***'* ******* ******** **********

***** ****** ******** ** the******* ******* ************* ******** **** *********** increments (*.*. **.*°*). *******, Ebrahimi **** ****** **** only **** ******* ******* has ** "*****-*******" **** temperature, ** * ******* degree ** **********:

** ****** ** *** that ** **** ** putting * ****** ****** someone's **** *** ****** 101°F; **** ***** **** you **** ********** *********** [to **** **** *** temperature] ** ***°*, *** we *** *** ***** down **** ****. ** are ***** **** *** road **: **** **** person **** * *********** above *** ***** ** everybody **** ******* ** in ***** ** **** camera? ** *****'* **** we *** ****** **** the ***********, **** **** this ****** *** * higher *********** **** *******. We ***** *** * probability ********* **** ******* has * ****** *********** than ******* - ** it's **% ****'* *** that **** ** **%.

**** ******** ***** *********** hurdles. *******, *** *****'* top ********* ************ *** electrotechnical ********, *** ***,************ ********** ***** **** **** display ********* **** *********** in *.*°* **********:

IPVM Image

*** ** ************* ************** ***** ********* ** well. ****. ******** *** not ***** ** **** requirement **** ** ******* it **.

Effective ******** ** * ******

******'* ******* ********* ******** is * ****** (** feet), ******** ******* *** not *** ***** ******* this **** ******** ** accurate ******, ******** ****.

**** ** * **** long ******** *** ***** cams *** **** ** skeptical **** ****** ***** ever ****** * ****** result. **** ** *** example **** *** *** considers * "******** ********** individual" - ******* *** distance ** **** * few **** **** **** much ****** **********:

IPVM Image

Camera **** "*****" ** *******, ******* *******

** ***** ** **** mass ******** ** **** as ********, ****** ** slated ** ** ******* powered/rechargeable *** *** *** and **** ******* ** mounting/screws/etc, ******** **** ****.

******: *** ***** **** be * "****** ******" (see *****), **** ** "external ***** **** *** be ****** ******* ** glued ** * ****, a ****** ** * mount ** * ******," Ebrahimi ****. *** ******* would **** "* ** 3 ******".

IPVM Image

Long **** ** **********

**** *********** *** ******* but ****** ***** * long **** ** **********:

  • Accuracy. The big question for ProCam is whether it can actually obtain accurate fever estimations for people or not. IPVM does not believe this is possible via a combination of cheap sensors and AI.
  • **********.****** ***'* ******* ******** temperatures *** ****** *** only ******* **** "***** average". **** **** ******* to *********** ********* *** fever ****** *********, ***** explicitly ******* ******** *********** displays *** **** ******, meaning ****** ***** **** itself ** * ******* tricky *********.
  • **************.*** **** ** ****** is ** ***** ***** cameras ** *** ******, however, ** **** ************* ***********, ***** ********* ** such * *********** *********/******* task **** **'* ****** likely ******* ********* *********** in ****** ** **** a ****** ***** **** to ******** ************** ******, false *********, *********** **********, etc.
  • AI *** **** *******. ProCam's use of AI is unprecedented as IPVM is not aware of any studies or existing products which use algorithms to the extent ProCam does. Ebrahimi was unable to provide such evidence too. Perhaps this AI will work, however, ProCam would be a first.

Conclusion

*** *** ******* *****, IPVM ** ********* ** ProCam, *** ** *** happy ** **** ** when * ********* ***** out. **********, *** ***** in **** ********* ** producing *********** *** ********** and ** ** ***** watching ******* ****** (** someone ****) *****, ******* our **********, ******** ***** a ***** *** ******** fever *** *** *** masses.

Comments (2)

***********! * ** **** a ******* *** ***, heck **** ****, ***** ago * ***** **** never ******* * ***** have ******** ****** ****** in ** *** **** that ********** **** ************ models *** * ******** of *** *****.

Agree: 3
Disagree
Informative
Unhelpful
Funny

***** *** * *** of ***** ********** ********** of *** **** *****. The **** ***** **** this ******* **** ** AI ***** ***** ** sharpen *** ********. ***** as * ***** ** a ****. *** **** else, **** ** *** sensor.*****://***.******-*******.***/*******-***********/*******/********/***-**-*****-****-***-*******-******-***-****-****-***

Agree
Disagree: 1
Informative: 1
Unhelpful
Funny
Read this IPVM report for free.

This article is part of IPVM's 6,958 reports, 927 tests and is only available to members. To get a one-time preview of our work, enter your work email to access the full article.

Already a member? Login here | Join now
Loading Related Reports