YOLOv3: Training and inference in PyTorch
Switch branches/tags
Nothing to show
Clone or download
Failed to load latest commit information.
.github/ISSUE_TEMPLATE Update issue templates Sep 1, 2018
cfg add yolov3-spp.cfg Sep 21, 2018
checkpoints updates Aug 26, 2018
data updates Sep 2, 2018
utils BCE to CE lconf + batch size 16 Oct 11, 2018
.gitignore updates Sep 1, 2018
LICENSE Initial commit Aug 26, 2018
README.md updates Sep 4, 2018
detect.py updates Sep 24, 2018
models.py BCE to CE lconf + batch size 16 Oct 11, 2018
requirements.txt Initial commit Aug 26, 2018
test.py per-class mAP report Oct 10, 2018
train.py -batch_size from 12 to 16 Oct 15, 2018



This directory contains software developed by Ultralytics LLC, and is freely available for redistribution under the GPL-3.0 license. For more information on Ultralytics projects please visit: http://www.ultralytics.com


The https://github.com/ultralytics/yolov3 repo contains inference and training code for YOLOv3 in PyTorch. Training is done on the COCO dataset by default: https://cocodataset.org/#home. Credit to Joseph Redmon for YOLO (https://pjreddie.com/darknet/yolo/) and to Erik Lindernoren for the PyTorch implementation this work is based on (https://github.com/eriklindernoren/PyTorch-YOLOv3).


Python 3.6 or later with the following pip3 install -U -r requirements.txt packages:

  • numpy
  • torch
  • opencv-python


Start Training: Run train.py to begin training after downloading COCO data with data/get_coco_dataset.sh and specifying COCO path on line 37 (local) or line 39 (cloud).

Resume Training: Run train.py -resume 1 to resume training from the most recently saved checkpoint latest.pt.

Each epoch trains on 120,000 images from the train and validate COCO sets, and tests on 5000 images from the COCO validate set. An Nvidia GTX 1080 Ti will process about 10-15 epochs/day depending on image size and augmentation (13 epochs/day at 416 pixels with default augmentation). Loss plots for the bounding boxes, objectness and class confidence should appear similar to results shown here (results in progress to 160 epochs, will update).


Image Augmentation

datasets.py applies random OpenCV-powered (https://opencv.org/) augmentation to the input images in accordance with the following specifications. Augmentation is applied only during training, not during inference. Bounding boxes are automatically tracked and updated with the images. 416 x 416 examples pictured below.

Augmentation Description
Translation +/- 20% (vertical and horizontal)
Rotation +/- 5 degrees
Shear +/- 3 degrees (vertical and horizontal)
Scale +/- 20%
Reflection 50% probability (horizontal-only)
HSV Saturation +/- 50%
HSV Intensity +/- 50%



Checkpoints are saved in /checkpoints directory. Run detect.py to apply trained weights to an image, such as zidane.jpg from the data/samples folder, shown here. Alternatively you can use the official YOLOv3 weights:



Run test.py to validate the official YOLOv3 weights checkpoints/yolov3.weights against the 5000 validation images. You should obtain a mAP of .581 using this repo (https://github.com/ultralytics/yolov3), compared to .579 as reported in darknet (https://arxiv.org/abs/1804.02767).

Run test.py -weights_path checkpoints/latest.pt to validate against the latest training checkpoint.


For questions or comments please contact Glenn Jocher at glenn.jocher@ultralytics.com or visit us at http://www.ultralytics.com/contact