Skip to content

sonalrpatel/object-detection-yolo

Repository files navigation

object-detection-yolo

Introduction

This repository is an implementation of YOLOv3 in Tensorflow. It contains complete pipeline for training and prediction on custom datasets.
The primary references are yolo3-tf2 and keras-yolo3 which helped me to understand the implementations and bring up this repo.

Predict

convert pre-trained Darknet weights to h5 format

# yolov3
wget https://pjreddie.com/media/files/yolov3.weights -O data/yolov3.weights
python convert.py --weights_path data/yolov3.weights --output_path data/yolov3_coco.h5

detection on a single image

# yolov3
python predict.py -w data/yolov3_coco.h5 -c data/coco_classes.txt -i data/sample/apple.jpg

Training

pending to update

Benchmark / Result

pending to update

References

The various resources referred are organised and listed below.

YOLOv1

YOLOv2

model build - data loading - train - predict

YOLOv3

Implementation in pyTorch

model build - data loading - train - predict

Implementation in Tensorflow

pre trained model - predict

model build - data loading - train - predict

custom training

transfer learning with yolo

load yolo weights

YOLOv4

load yolo weights

YOLOv5

Model Subclassing

Custom Loss

Dataset / Format

Todo

  • Update the training and benchmark / result sections

About

object detection using yolo architecture

Resources

Stars

Watchers

Forks

Packages

No packages published