Skip to content
master
Go to file
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 

README.md

Aerial Object Detector

An object detection system for aerial data (esp. for DOTA dataset)

using Tensorflow Object Detection API

Competition

Applications

Dataset

Object Detection

Mask-RCNN approach

Faster-RCNN approach

YOLO-based approach

Convert DOTA to YOLO (Darknet) format

In DOTA, the annotation format is:

    x1 y1 x2 y2 x3 y3 x4 y4 category difficult

In YOLO (Darknet), the below annotation format is required

    category-id x y width height

To do so, run the script below

python convert_to_darknet.py

Train

python /data/private/models/research/object_detection/model_main.py --pipeline_config_path='./configs/faster_rcnn_resnet101_dota.config' --train_dir=./checkpoints/faster_rcnn_dota --num_train_steps=1000 --alsologtostderr

Download model train weights:

wget https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md
tar -xzvf faster_rcnn_resnet101_coco_2018_01_28.tar.gz

About

To develop a object detection system for aerial data

Resources

Releases

No releases published

Packages

No packages published

Languages