There is a newer version of this record available.

Software Open Access

ultralytics/yolov5: v3.1 - Bug Fixes and Performance Improvements

Glenn Jocher; Alex Stoken; Jirka Borovec; NanoCode012; ChristopherSTAN; Liu Changyu; Laughing; tkianai; Adam Hogan; lorenzomammana; yxNONG; AlexWang1900; Laurentiu Diaconu; Marc; wanghaoyang0106; ml5ah; Doug; Francisco Ingham; Frederik; Guilhen; Hatovix; Jake Poznanski; Jiacong Fang; Lijun Yu 于力军; changyu98; Mingyu Wang; Naman Gupta; Osama Akhtar; PetrDvoracek; Prashant Rai

Dublin Core Export

<?xml version='1.0' encoding='utf-8'?>
<oai_dc:dc xmlns:dc="" xmlns:oai_dc="" xmlns:xsi="" xsi:schemaLocation="">
  <dc:creator>Glenn Jocher</dc:creator>
  <dc:creator>Alex Stoken</dc:creator>
  <dc:creator>Jirka Borovec</dc:creator>
  <dc:creator>Liu Changyu</dc:creator>
  <dc:creator>Adam Hogan</dc:creator>
  <dc:creator>Laurentiu Diaconu</dc:creator>
  <dc:creator>Francisco Ingham</dc:creator>
  <dc:creator>Jake Poznanski</dc:creator>
  <dc:creator>Jiacong Fang</dc:creator>
  <dc:creator>Lijun Yu 于力军</dc:creator>
  <dc:creator>Mingyu Wang</dc:creator>
  <dc:creator>Naman Gupta</dc:creator>
  <dc:creator>Osama Akhtar</dc:creator>
  <dc:creator>Prashant Rai</dc:creator>
  <dc:description>This release aggregates various minor bug fixes and performance improvements since the main v3.0 release and incorporates PyTorch 1.7.0 compatibility updates. v3.1 models share weights with v3.0 models but contain minor module updates (inplace fields for nn.Hardswish() activations) for native PyTorch 1.7.0 compatibility.
Breaking Changes

'GIoU' hyperparameter has been renamed to 'box' to better reflect a criteria-agnostic regression loss term  (

Bug Fixes

PyTorch 1.7 compatibility update. torch&gt;=1.6.0 required, torch&gt;=1.7.0 recommended (
GhostConv module bug fix (
Rectangular padding min stride bug fix from 64 to 32 (
Mosaic4 bug fix (
Logging directory runs/exp bug fix (
Various additional

Added Functionality

PyTorch Hub functionality with YOLOv5 .autoshape() method added (
Autolabelling addition and standardization across and (
Precision-Recall Curve automatic plotting when testing (
Self-host VOC dataset for more reliable access and faster downloading (
Adding option to output autolabel confidence with --save-conf in and (
Google App Engine deployment option (
Infinite Dataloader for faster training (
Various additional
  <dc:title>ultralytics/yolov5: v3.1 - Bug Fixes and Performance Improvements</dc:title>
All versions This version
Views 12,5863,256
Downloads 22046
Data volume 504.9 MB146.7 MB
Unique views 10,0032,664
Unique downloads 18739


Cite as