韩国网页设计公司网站,湖北系统app定制开发系统,建企业网站公司,如何对网站进行改版关于Grounding DINO的环境搭建可以参考我的以前的博客#xff0c;链接如下所示
如何在Linux上离线部署Grounding DINO-CSDN博客
这个博客主要来介绍如何利用Grounding DINO这个项目去进行目标检测的自动化标注。并且给出了相关的代码已经实验验证。
1.数据集准备 2. 开始实…
关于Grounding DINO的环境搭建可以参考我的以前的博客链接如下所示
如何在Linux上离线部署Grounding DINO-CSDN博客
这个博客主要来介绍如何利用Grounding DINO这个项目去进行目标检测的自动化标注。并且给出了相关的代码已经实验验证。
1.数据集准备 2. 开始实验
2.1 批量标注参考代码如下
import os
import cv2
import torch
from torchvision.ops import box_convert
from groundingdino.util.inference import load_model, load_image, predict, annotate# 配置路径
MODEL_CONFIG_PATH groundingdino/config/GroundingDINO_SwinT_OGC.py
MODEL_WEIGHTS_PATH weights/groundingdino_swint_ogc.pth
PROJECT_ROOTAuto_label/Project1/ # 自动检测的根路径
IMAGE_FOLDER PROJECT_ROOT images # 输入图片文件夹
OUTPUT_FOLDER PROJECT_ROOT detect_results # 输出标注图片的文件夹
LABELS_FOLDER PROJECT_ROOT labels # 输出YOLO标签的文件夹
CLASSES_FILE PROJECT_ROOT classes.txt # 类别文件# YOLO标签格式转换函数
def convert_to_yolo_format(xyxy, image_width, image_height):将 xyxy 坐标转换为 YOLO 格式的 x_center, y_center, width, heightx_min, y_min, x_max, y_max xyxyx_center (x_min x_max) / 2.0 / image_widthy_center (y_min y_max) / 2.0 / image_heightwidth abs(x_max - x_min) / image_widthheight abs(y_max - y_min) / image_heightreturn x_center, y_center, width, height# 加载类别文件
def load_classes(classes_file):with open(classes_file, r) as f:return [line.strip() for line in f.readlines()]# 主检测与标签生成函数
def process_images(model, classes, image_folder, output_folder, labels_folder):os.makedirs(output_folder, exist_okTrue)os.makedirs(labels_folder, exist_okTrue)for image_file in os.listdir(image_folder):if not image_file.lower().endswith((.png, .jpg, .jpeg)):continue# 加载图片image_path os.path.join(image_folder, image_file)image_source, image load_image(image_path)h, w, _ image_source.shape# 推理检测boxes, logits, phrases predict(modelmodel,imageimage,# captioncar . coach . bus . truck . tricycle . person . twowheelsvehicle . taxi . license_plate . other_vehicles,captionperson . car . dog . cat,box_threshold0.35, #0.35text_threshold0.25 # 0.25)# 缩放坐标并转换为 xyxyyolo_boxes boxes * torch.Tensor([w, h, w, h])xyxy_boxes box_convert(boxesyolo_boxes, in_fmtcxcywh, out_fmtxyxy).numpy()# 创建YOLO标签文件label_file os.path.join(labels_folder, os.path.splitext(image_file)[0] .txt)with open(label_file, w) as label_f:for xyxy, phrase in zip(xyxy_boxes, phrases):# 获取类别索引class_idx classes.index(phrase) if phrase in classes else -1if class_idx -1:continue # 跳过不在类别文件中的目标# 转换坐标格式x_center, y_center, width, height convert_to_yolo_format(xyxy, w, h)# 写入YOLO标签文件label_f.write(f{class_idx} {x_center:.6f} {y_center:.6f} {width:.6f} {height:.6f}\n)# 标注图片并保存annotated_frame annotate(image_sourceimage_source, boxesboxes, logitslogits, phrasesphrases)output_image_path os.path.join(output_folder, image_file)cv2.imwrite(output_image_path, annotated_frame)print(fProcessed {image_file}, labels saved to {label_file}, annotated image saved to {output_image_path})# 主函数
if __name__ __main__:# 加载模型和类别model load_model(MODEL_CONFIG_PATH, MODEL_WEIGHTS_PATH)classes load_classes(CLASSES_FILE)# 处理图片并生成标签process_images(model, classes, IMAGE_FOLDER, OUTPUT_FOLDER, LABELS_FOLDER)上面的注释非常详细了就不过多赘述了。
主要根据自己的环境修改以下内容
配置路径提示词 2.2 开始实验
我主要想检测因此我的提升词设置如下 captionperson . car . dog . cat, 检测结果
Auto_label/Project1/detect_results文件夹 标签文件
Auto_label/Project1/labels文件夹 2.3 实验验证
分析获得标签是否正确可以可视化标签可视化标签代码可以参考我以前的博客链接如下所示
目标检测-可视化YOLO格式标签_yolo标签可视化-CSDN博客 非常的完美 最近有更新了Grounding DINO-X其效果更好链接如下
GitHub - IDEA-Research/DINO-X-API: DINO-X: The Worlds Top-Performing Vision Model for Open-World Object Detection and Understanding
