YOLO in pytorch

introduction:

this is an implementation of YOLO object detection algorithme. the implementation is a mix between YOLO v1 and v2. I used pytorch models trained on ImageNet as backbone model then added final conv layer for the output THERE IS NO FULLY CONNECTED LAYERS

Dependencies:

torch==1.3.1
torchvision==0.4.2
opencv-python==4.1.2.30
albumentations==0.4.3
numba==0.46.0

installation:

clone and install the requirements

git clone https://github.com/AhmedGhazale/cifar100-classifier.git
cd cifar100-classifier
pip3 install -r requirements.txt

Demo :

to run an image

python3 predict.py path/to/image

to run video

python3 video_demo.py path/to/video

the output will be a video named output.avi in the same directory

Traininig:

download voc dataset
edit the dataset path in config.py
run

python3 train.py

evaluation:

TODO

AhmedGhazale / YOLO_pytorch