YOLOv5 Released Amidst Controversy

By Zach Segal, Published Jul 27, 2020, 11:11am EDT

YOLO has gained significant attention within video surveillance for its promise of better, faster video analytics. However, 2020 has proven to be controversial for the platform.

IPVM Image

Inside this report, we:

  • Explain what happened to the Original Team Including XNOR.ai
  • Describe the Future of YOLO Development
  • YOLOv5 Controversy
  • YOLOv5 vs YOLOv4 vs YOLOv3
  • Real-World Impact

Executive *******:

** ****, ****** *******, along **** *** ******* and * *** ******, released ****** (*** **** Look **** ******* *), a ***** ****** ********* algorithm **** *** **** and ********. ****, ***** releasing ******** * *** 3,********** *** ******** **** privacy ***************.**, ***** *** ******* by *** ***** ******* of *** ******** **** team, *********** ** ************* ********* Apple, ******* *** ***** of **** ** ****.

****, ****** *********** **************, ***** *** ************ improvements **** **, *** was ****** ********* ** the **** **** ** the **** ******* ************ *** ******** ** large.****, **** ***** ******/*********** released******, ** *** ********* controversial ******* *** ************ were ** **** ** use (*** ******** ** speed) **** **** (********* Bochovskiy) ******* ** ** nothing **** **** * tweaked ******* ** ******. Also, ****** *** ******** versions, ***** *** ** paper ******** ** ** (so ***** *** ** peer-review *******) *** ** was ********* **** ***** to * ******* *******. Regardless, ****** ***/** ****** have *** ********* ** make ****-**** ****** ********* more ************ ********.

What ** ****

**** ** *** **** Look **** ****** *** a ***** ** ****** detection ********** **** *** to ** ****** ********* accurately *** ***********. **** introduced *** **** ** using *** ****** ******* for ******** *** *** class **********, ******* * two-step ******* **** *** piece, ******** ****** ************. They **** ***** ********** in**** ** * ******** student,****** ******, *** **** garnered **** ** *********. He ******** *** **** updated ********, ****** *** YOLOv3; ****** ****** *** research *** ******* *** industry.

Why ******* **** ****

******* *****'******** ************ *** ******* concerns' ** ******* ** stopped:

IPVM Image

XNOR.ai ** *****

*** **** ** *** original **** **** ******* XNOR.ai,***** *** ****** ** Apple *** $*** *********** ****. *** ** Apple’s ******* ***** ** privacy *** ***-**-*** ********, its ****** *&* **** probably *** ** *********. Indeed, ****.** ********** **** and ******* ******** *** analytics ** ***** *********.

The ***** ** **** *****

**** **** **** *** seen *** *** ***** create***************. *******, ***** ****** Bochkovskiy’s ****** *********************** ********* ******** ******, ***** *** ******************* ***** ******/**********’* ******.

The ***********

****** ** *** ****** an *********** **** ******. YOLOv5 ** ****** *** less ******** **** ******, but **** **** ********. Specifically, ****** ******** **** versions **** ******** *******, YOLOv5 **** ******** ** PyTorch, * ****** **** deep ******** *********.

**** ********* ***** ******* why ** ** ***** used ************ **************** (******* **** ******* contests).

IPVM Image

****, ****** ******** ******, ****** was ******** *********-********* ** a ******* *******, ******* a ****-******** ***** ******** to **. **** ***** the ****** **** ****** made **** *** ******, there ** ********* **** when *** **** ******* will ** **** *** what **** **** **** like, *** ** *** be **** ** * PR **** ******* ** an ******* ** ******* computer ******.

Improvements ******* ** *** **

****** *** *********** ************ over ******** ***** ** ***** and ********. ** *** MS **** ****** ********* dataset, ** ** *** YOLOv4 ******** ~**% ****** AP50 (~** **. ~**) than ******.

IPVM Image

YOLOv4 *****

*********** ***** *** ******, ********* *** **** accurate *** ******* ******* are ******.**% **** **** ~** fps ** ** ****** 1080Ti (~$***).

** ********, ******-**** *** significantly ******* **** ** 40.2% ***** **** **** 8x ******** ** *** at ~*** *** **** the **** **** **(~$***).

IPVM Image

Potential ***** ************ ************

**** *** **** *** years, * ****** ** video ************ ********* **** used ** ******* ********** of **** ** *** in ********** ***** *********. While **** ******** ***** on ** ********** (******* **** ********), *** *********** *** video ************ ************ *** be ******* ********* *********** categories **** ***** ************ users *** ** **** interested (*.*., *** * categories ** *******).

** **** ***, ********** of ****** ***** ******* further ************ *** ***** surveillance *** *** ******* over *** **** **** within **** *** ****** the ********* *** ****** gains **** **** ********.

Comments (5)

Greg Nuckles

IPVMU Certified | 07/28/20 04:26pm

**** ****!

Zach Segal

IPVM | In reply to Greg Nuckles | 07/28/20 04:54pm

***** ***

Avatar

Igor Falomkin

AxxonSoft | 07/28/20 04:53pm

** (*********) **** ***** YOLO4 *** *** ****** detection ** ********* ********** and ******** ** **** Mobilenet ** *** ********. All ***** ******** **** trained ** *** *** and ****** ** *** same **** **** **** CCTV ******. ********* (********* false ********* ****) *** nearly *** **** *** all ********, ****** (********* detection ****) ** ****** the **** *** ***** and ********, *** **** are **** ****** **** Mobilenet. ******** ** ****** 3 ***** "******" **** Mobilenet, ***** ** ** times ****** **** *********. As * ****** ** decided **** *** *** goals ******** ** ****** than *****. ** ***** is *** *** **** choice *** ****.

Zach Segal

IPVM | In reply to Igor Falomkin | 07/28/20 05:04pm

***** *** *** ******** this **. **'* ***** to **** **** ********** have **** **** *** in ******** *** ***** and *** **** **** picked.

Zach Segal

IPVM | 08/06/20 09:43pm

* **** **** ***** has ************* * ******* **-**** ***** *********** YOLOv4. ** **** ** interesting ** *** ***/** the ********** **** ***** programs *** ******** ** make ****** *** **** accurate ****** ********* ******.

Read this IPVM report for free.

This article is part of IPVM's 6,815 reports, 914 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