Paddle-Anti-UAV

Anti-UAV base on PaddleDetection

Background

UAVs are very popular and we can see them in many public spaces, such as parks and playgrounds. Most people use UAVs for taking photos. However, many areas like airport forbiden UAVs since they are potentially dangerous. In this case, we need to detect the flying UAVs in these areas.

In this repository, we show how to train a detection model using PaddleDetection.

Data preparation

The dataset can be found here. We direcly download the test-dev split composed of 140 videos train the detection model.

Download the test-dev dataset.
Run unzip Anti_UAV_test_dev.zip -d Anti_UAV.
Run python get_image_label.py. In this step, you may change the path to the videos and the value of interval.

After the above steps, you will get a MSCOCO-style datasst for object detection.

Install PaddleDetection

Please refer to this link.

We use python=3.7, Paddle=2.2.1, CUDA=10.2.

Train PP-YOLO

We use PP-YOLO as the detector.

Run git clone https://github.com/PaddlePaddle/PaddleDetection.git. Note that you should finish this step when you install PaddleDetection.
Move the anti-UAV dataset to dataset.
Move anti_uav.yml to configs/datasets, move ppyolo_r50vd_dcn_1x_antiuav.yml to configs/ppyolo and move ppyolo_r50vd_dcn_antiuav.yml to configs/ppyolo/_base.
Keep the value of anchors in configs/ppyolo/_base/ppyolo_reader.yml the same as ppyolo_r50vd_dcn_antiuav.yml.
Run python -m paddle.distributed.launch --log_dir=./ppyolo_dygraph/ --gpus 0,1,2,3,4,5,6,7 tools/train.py -c configs/ppyolo/ppyolo_r50vd_dcn_1x_antiuav.yml &>ppyolo_dygraph.log 2>&1 &. Note that you may change the arguments, such as batch_size and gups.

Inference

Please refer to the infernce section on this webpage. You can just switch the configeration file and trained model to your own files.

Anti-UAV base on PaddleDetection

Related tags

Overview

Paddle-Anti-UAV

Background

Data preparation

Install PaddleDetection

Train PP-YOLO

Inference

Owner

Qingzhong Wang

git《Self-Attention Attribution: Interpreting Information Interactions Inside Transformer》(AAAI 2021) GitHub:

O-CNN: Octree-based Convolutional Neural Networks for 3D Shape Analysis

High-resolution networks and Segmentation Transformer for Semantic Segmentation

Code for the ICASSP-2021 paper: Continuous Speech Separation with Conformer.

Converts geometry node attributes to built-in attributes

A framework for using LSTMs to detect anomalies in multivariate time series data. Includes spacecraft anomaly data and experiments from the Mars Science Laboratory and SMAP missions.

PyTorch implementation of Octave Convolution with pre-trained Oct-ResNet and Oct-MobileNet models

Implementation of ResMLP, an all MLP solution to image classification, in Pytorch

Bridging the Gap between Label- and Reference based Synthesis(ICCV 2021)

Low-code/No-code approach for deep learning inference on devices

Official code repository for the EMNLP 2021 paper

Code that accompanies the paper Semi-supervised Deep Kernel Learning: Regression with Unlabeled Data by Minimizing Predictive Variance

Norm-based Analysis of Transformer

The Habitat-Matterport 3D Research Dataset - the largest-ever dataset of 3D indoor spaces.

An Intelligent Self-driving Truck System For Highway Transportation

Fashion Entity Classification

Multi-Modal Machine Learning toolkit based on PaddlePaddle.

This is the repository for Learning to Generate Piano Music With Sustain Pedals

《Train in Germany, Test in The USA: Making 3D Object Detectors Generalize》(CVPR 2020)

StudioGAN is a Pytorch library providing implementations of representative Generative Adversarial Networks (GANs) for conditional/unconditional image generation.