1. Three detection heads ASFF ①common.py Add the following code at the bottom of the models/common.py file: def add_conv(in_ch, out_ch, ksize, stride, leaky=True): “”” Add a conv2d / batchnorm / leaky ReLU block. Args: in_ch (int): number of input channels of the convolution layer. out_ch (int): number of output channels of the convolution layer. ksize […]
Tag: asf
YOLOv5, YOLOv8 add ASFF (Adaptive Spatial Feature Fusion)
ASFF: Adaptively Spatial Feature Fusion (adaptive spatial feature fusion) Paper source: Learning Spatial Fusion for Single-Shot Object Detection Code address: ASFF 1.Background Inconsistency between differentfeature scalesis a major drawback of single-stage detectors based on feature pyramids. This paper proposes a new data-driven pyramid feature fusion strategy called adaptive spatial feature fusion (ASFF). It learns a […]
The relationship and implementation principle of Endpoint, HasFeatures, NamedFeature and Actuator in Spring
Article directory 1. Reason for relationship 2. Introduction and simple use of Actuator 3. The relationship between Endpoint and Actuator 4. The relationship between Endpoint and HasFeatures 5. Analysis of Endpoint and HasFeatures principles 5.1 Implementation principle of Endpoint 5.2 Implementation principle of HasFeatures 6. Personal chat 1. Reason for relationship We can often see […]
[Yolov5] Yolov5 adds ASFF and a new detection layer at the same time
Yolov5 adds ASFF and new detection layer at the same time Foreword The blogger learned Yolov5 some time ago in order to do graduation design. Now that the tasks required for graduation have been completed, he has not continued to study the improvement method of Yolov5. However, I saw that some friends in the comment […]
A scenario of using Jackson-AnnotationIntrospector + @AliasFor to implement custom annotation + JSON serialization in SpringBoot
Table of Contents foreword train of thought accomplish 1. Custom annotations 2. Customize JacksonAnnotationIntrospector 1.2.1 General situation 1.2.2 Implementation of this example 3. Configuration 4. Example of use Summarize Foreword Requirements: dynamic header (relatively speaking) At present, only simple operations such as controlling column width and column name, sorting, displaying and hiding are not available. […]
YOLOv5 improvement adds decoupling head, ASFF
There are many blogs on the Internet that add decoupling headers. I record here that I use decoupling headers to improve YOLOv5. The introduction of the decoupling header will be written after a while, and the method will be added directly I have not used ASFF, but it can also run after adding it according […]
Road defect recognition based on Yolov5, adding ASFF optimization
1. Dataset introduction Defect type: crack Number of datasets: 195 1.1 Data enhancement, expanding the dataset Through medianBlur, GaussianBlur, Blur 3 times expansion to get 780 pictures 8:1:1 division according to train, val, test 1.1.1 Get trainval.txt, val.txt, test.txt through split_train_val.py # coding:utf-8 import os import random import argparse parser = argparse. ArgumentParser() #xml file […]
Tips for increasing points: Detect series—Yolov5/Yolov7 joins the ASFF feature pyramid fusion method, and the rising point is obvious
Directory 1. Introduction to ASFF 2. ASFF joins Yolov5 to improve detection accuracy 2.1 Add ASFF to common.py: 2.2 Add ASFF to yolo.py: 2.3 Modify yolov5s_asff.yaml 2.4 Combining with cbam to further improve detection accuracy 1. Introduction to ASFF Learning Spatial Fusion for Single-Shot Object Detection Paper address: https://arxiv.org/pdf/1911.09516v2.pdf Multi-scale features, especially feature pyramid FPN, […]