EADST

YOLOv5: Data Preparation for Training

YOLOv5: Data Preparation for Training, from VOC Format to YOLO Format

First, split the data to train and val, and get txt files

import os
import random

train_percent = 0.95 xmlfilepath = 'bucket_v1/Annotations' txtsavepath = 'bucket_v1/ImageSets' total_xml = os.listdir(xmlfilepath)

num = len(total_xml) trainval = range(num) tr = int(num * train_percent) train = random.sample(trainval, tr)

ftrain = open('bucket_v1/ImageSets/Main/train.txt', 'w') fval = open('bucket_v1/ImageSets/Main/val.txt', 'w')

for i in trainval: name = total_xml[i][:-4] + '\n' if i in train: ftrain.write(name) else: fval.write(name)

Second, convert VOC format data (xml) to YOLO format data (txt)

import os
from tqdm import tqdm
from lxml import etree
import json
import shutil

voc_root = "/dfs/data/others/byolov5/dataset/bucket_v1" voc_version = "bucket_v1"

train_txt = "train.txt" val_txt = "val.txt" save_file_root = "/dfs/data/others/byolov5/dataset/yolo_data"

voc_images_path = os.path.join(voc_root, "JPEGImages") voc_xml_path = os.path.join(voc_root, "Annotations") train_txt_path = os.path.join(voc_root, "ImageSets", "Main", train_txt) val_txt_path = os.path.join(voc_root, "ImageSets", "Main", val_txt)

def parse_xml_to_dict(xml): if len(xml) == 0: return {xml.tag: xml.text} result = {} for child in xml: child_result = parse_xml_to_dict(child)
if child.tag != 'object': result[child.tag] = child_result[child.tag] else: if child.tag not in result:
result[child.tag] = [] result[child.tag].append(child_result[child.tag]) return {xml.tag: result}

def translate_info(file_names: list, save_root: str, class_dict: dict, train_val='train'): save_txt_path = os.path.join(save_root, train_val, "labels") if os.path.exists(save_txt_path) is False: os.makedirs(save_txt_path) save_images_path = os.path.join(save_root, train_val, "images") if os.path.exists(save_images_path) is False: os.makedirs(save_images_path)

for file in tqdm(file_names, desc="translate {} file...".format(train_val)):
    img_path = os.path.join(voc_images_path, file + ".jpg")
    assert os.path.exists(img_path), "file:{} not exist...".format(img_path)

    xml_path = os.path.join(voc_xml_path, file + ".xml")
    assert os.path.exists(xml_path), "file:{} not exist...".format(xml_path)

    # read xml
    with open(xml_path, encoding='UTF-8') as fid:
        xml_str = fid.read()
    xml = etree.fromstring(xml_str)
    data = parse_xml_to_dict(xml)["annotation"]
    img_height = int(data["size"]["height"])
    img_width = int(data["size"]["width"])
    # write object info into txt
    # assert "object" in data.keys(), "file: '{}' lack of object key.".format(xml_path)
    if "object" not in data.keys():
        print("Warning: in '{}' xml, there are no objects.".format(xml_path))
        continue

    with open(os.path.join(save_txt_path, file + ".txt"), "w") as f:
        for index, obj in enumerate(data["object"]):
            xmin = float(obj["bndbox"]["xmin"])
            xmax = float(obj["bndbox"]["xmax"])
            ymin = float(obj["bndbox"]["ymin"])
            ymax = float(obj["bndbox"]["ymax"])
            class_name = obj["name"]
            class_index = class_dict[class_name] - 1

            if xmax <= xmin or ymax <= ymin:
                print("Warning: in '{}' xml, there are some bbox w/h <=0".format(xml_path))
                continue

            xcenter = xmin + (xmax - xmin) / 2
            ycenter = ymin + (ymax - ymin) / 2
            w = xmax - xmin
            h = ymax - ymin

            xcenter = round(xcenter / img_width, 6)
            ycenter = round(ycenter / img_height, 6)
            w = round(w / img_width, 6)
            h = round(h / img_height, 6)

            info = [str(i) for i in [class_index, xcenter, ycenter, w, h]]

            if index == 0:
                f.write(" ".join(info))
            else:
                f.write("\n" + " ".join(info))

    # copy image into save_images_path
    path_copy_to = os.path.join(save_images_path, img_path.split(os.sep)[-1])
    if os.path.exists(path_copy_to) is False:
        shutil.copyfile(img_path, path_copy_to)

def main(): class_dict = {"b": 1, "t": 2} with open(train_txt_path, "r") as r: train_file_names = [i for i in r.read().splitlines() if len(i.strip()) > 0] translate_info(train_file_names, save_file_root, class_dict, "train") with open(val_txt_path, "r") as r: val_file_names = [i for i in r.read().splitlines() if len(i.strip()) > 0] translate_info(val_file_names, save_file_root, class_dict, "val")

if name == "main": main()

Reference:

Train Custom Data

相关标签
About Me
XD
Goals determine what you are going to be.
Category
标签云
SVR git OpenAI Heatmap UNIX 版权 算法题 Algorithm 证件照 IndexTTS2 Base64 OpenCV 图形思考法 Paddle BeautifulSoup Pytorch Plotly Interview InvalidArgumentError Github Input Statistics Excel Gemma Food Domain Knowledge mmap BTC GoogLeNet JSON PyCharm Harness 长期记忆 Web Vim Hilton Tensor DeepSeek Review GPT4 FP32 Google DeepStream NLP 个人博客 TensorRT Translation Shortcut Rebuttal Streamlit FlashAttention Bipartite 搞笑 tar News 净利润 Docker NLTK LLM git-lfs 强化学习 transformers TSV Augmentation 域名 uwsgi 飞书 printf Math PDF Image2Text Cloudreve LLAMA Search 自我改进 Distillation llama.cpp SPIE Clash TensorFlow FP16 Baidu tqdm FastAPI hf Proxy Animate Bin Ptyhon RGB Safetensors VSCode API 公式 Logo SQL Quantization PDB CTC API网关 CV 签证 LoRA EXCEL 财报 OCR v2ray Hungarian Datetime 音频 ONNX v0.dev Bert Video diffusers Tiktoken GIT C++ Nginx Ubuntu 阿里云 腾讯云 图标 Hotel 多进程 Paper ModelScope QWEN WAN Card SQLite Attention Color BF16 scipy GPTQ PyTorch AI RL ChatGPT Linux Bitcoin ms-swift GGML 论文速读 Template Crawler Plate Dataset LeetCode 顶会 SAM Magnet uWSGI Zip UI Conda ResNet-50 Quantize FP8 CLAP Qwen Permission Jev Password icon Breakpoint XGBoost Data YOLO Freesound HaggingFace HuggingFace Windows Sklearn 论文 Miniforge 多线程 Mixtral Claude Pickle Firewall 继承 torchinfo Jetson Pillow Use NameSilo Disk VPN Python Django LaTeX COCO 云服务器 递归学习法 PIP Llama Land CEIR Random RAR Numpy Pandas Diagram Transformers logger 关于博主 WebCrawler Markdown XML 第一性原理 VGG-16 Git CSV Website Qwen2.5 CUDA Michelin Anaconda FP64 报税 Vmess CAM Tracking Agent Jupyter MD5 Qwen2 TTS CC
站点统计

本站现有博文340篇,共被浏览1021348次

本站已经建立2719天!

热门文章
文章归档
回到顶部