https://github.com/avanetten/yoltv5

YOLT, now with PyTorch.
https://github.com/avanetten/yoltv5

Last synced: 12 days ago
JSON representation

YOLT, now with PyTorch.

• Host: GitHub
• URL: https://github.com/avanetten/yoltv5
• Owner: avanetten
• License: gpl-3.0
• Created: 2022-02-04T12:26:42.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2023-02-09T07:33:53.000Z (over 1 year ago)
• Last Synced: 2024-02-29T09:35:28.813Z (4 months ago)
• Language: Python
• Size: 1.9 MB
• Stars: 188
• Watchers: 8
• Forks: 26
• Open Issues: 14
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Lists

• awesome-yolo-object-detection - YOLTv5

README

        
# YOLTv5 #

![Alt text](/results/__examples/header.jpg?raw=true "")

 

 YOLTv5 rapidly detects objects in arbitrarily large aerial or satellite images that far exceed the ~600×600 pixel size typically ingested by deep learning object detection frameworks. 

  

 YOLTv5 builds upon [YOLT]( https://github.com/avanetten/yolt) and [SIMRDWN]( https://github.com/avanetten/simrdwn), and updates these frameworks to use the [YOLOv5](https://github.com/ultralytics/yolov5) version of the [YOLO](https://pjreddie.com/darknet/yolo/) object detection family.  This repository has generally similar performance to the [Darknet](https://pjreddie.com/darknet/)-based [YOLTv4](https://github.com/avanetten/yoltv4) repository.  For those users who prefer a [PyTorch](https://pytorch.org) backend, however, we provide YOLTv5.  

 

 Below, we provide examples of how to use this repository with the open-source [SpaceNet](https://spacenet.ai) dataset. 

 

____

## Running YOLTv5

___

### 0. Installation (Preliminary)

YOLTv5 is built to execute on a GPU-enabled machine. 

	cd yoltv5/yolov5

	pip install -r requirements.txt 

	# update with geo packages

	conda install -c conda-forge gdal

	conda install -c conda-forge osmnx=0.12 

	conda install  -c conda-forge scikit-image

	conda install  -c conda-forge statsmodels

	pip install torchsummary

	pip install utm

	pip install numba

	pip install jinja2==2.10

___

### 1. Train

Training preparation is accomplished via [prep_train.py](https://github.com/avanetten/yoltv5/blob/main/yoltv5/prep_train.py).  To train a model, run:

	cd /yoltv5

    python yolov5/train.py --img 640 --batch 16 --epochs 100 --data yoltv5_train_vehicles_8cat.yaml --weights yolov5l.pt

___

### 2. Test

Simply edit [yoltv5_test_vehicles_8cat.yaml](https://github.com/avanetten/yoltv5/blob/main/configs/yoltv5_test_vehicles_8cat.yaml) to point to the appropriate locations, then run the _test.sh_ script:

	cd yoltv5

	./test.sh ../configs/yoltv5_test_vehicles_8cat.yaml

Outputs will look something like the figure below (cars=green, trucks=red, buses=blue):

![Alt text](/results/__examples/khartoum_example0.jpg?raw=true "")