DAMO-YOLO训练自己的数据集，使用onnxruntime推理部署

DAMO-YOLO训练自己的数据集，使用onnxruntime推理部署

DAMO-YOLO 是阿里达摩院智能计算实验室开发的一种兼顾速度与精度的目标检测算法，在高精度的同时，保持了很高的推理速度。

DAMO-YOLO 是在 YOLO 框架基础上引入了一系列新技术，对整个检测框架进行了大幅的修改。具体包括：

1. 基于 NAS 搜索的新检测 backbone 结构，利用 MAE-NAS 方法快速搜索出适合检测任务的网络结构，比如MAE-NAS-L35 和 MAE-NAS-L45。
2. 更深的 neck 结构，采用 RepGFPN技术，实现了高效的多尺度特征融合，提升了特征表达能力和模型性能。 精简的 head 结构，采用 ZeroHead技术，减少了冗余的参数和计算量，提高了模型速度和精度。
3. 引入蒸馏技术实现效果的进一步提升，利用大模型作为教师模型，小模型作为学生模型，通过知识蒸馏方法提高小模型的泛化能力。
4. DAMO-YOLO还提供高效的训练策略和便捷易用的部署工具，能够快速解决工业落地中的实际问题。可以通过以下链接访问 DAMO-YOLO的代码和文档，或者通过以下链接访问 DAMO-YOLO 在 ModelScope 平台上的在线体验。

1.modelscope平台：https://modelscope.cn/models/damo/cv_tinynas_objectdetection_damoyolo/summary
2.github链接：https://github.com/tinyvision/DAMO-YOLO

1.数据准备

DAMO-YOLO目前支持coco数据集和voc数据集格式，本文推荐使用coco数据集格式，如果您是用labelimg标注的xml文件，可以使用下述voc2coco.py脚本进行一键转换（需要按我的格式摆放数据，否则的话需要改脚本中的路径）：

.
├── ./voc2coco.py
├── ./data
│   ├── ./data/Annotations
│   │   ├── ./data/Annotations/0.xml
│   │   ├── ./data/Annotations/1000.xml
│   │   ├── ...
│   ├── ./data/images
│   │   ├── ./data/images/0.jpg
│   │   ├── ./data/images/1000.jpg
│   │   ├── ...
1
2
3
4
5
6
7
8
9
10
11

voc2coco.py脚本，注意需要改下category_set类别（换成自己数据集的）：

import xml.etree.ElementTree as ET
import os
import json
import collections
import random
import shutil


category_set = ['ship']

# 设置随机数种子，可以是任意整数
random_seed = 42

# 设置随机数种子
random.seed(random_seed)


coco_train = dict()
coco_train['images'] = []
coco_train['type'] = 'instances'
coco_train['annotations'] = []
coco_train['categories'] = []

coco_val = dict()
coco_val['images'] = []
coco_val['type'] = 'instances'
coco_val['annotations'] = []
coco_val['categories'] = []

# category_set = dict()
image_set = set()
train_image_id = 1
val_image_id = 200000  # Assuming you have less than 200000 images
category_item_id = 1
annotation_id = 1


def split_list_by_ratio(input_list, ratio=0.8):
    # 计算切分的索引位置
    split_index = int(len(input_list) * ratio)
    # 随机打乱列表
    random.shuffle(input_list)
    # 划分为两个列表并返回
    return input_list[:split_index], input_list[split_index:]

def addCatItem(name):
    '''
    增加json格式中的categories部分
    '''
    global category_item_id
    category_item = collections.OrderedDict()
    category_item['supercategory'] = 'none'
    category_item['id'] = category_item_id
    category_item['name'] = name
    coco_train['categories'].append(category_item)
    coco_val['categories'].append(category_item)
    category_item_id += 1


def addImgItem(file_name, size, img_suffixes, is_train):
    global train_image_id  # 声明变量为全局变量
    global val_image_id  # 声明变量为全局变量
    # global image_id
    if file_name is None:
        raise Exception('Could not find filename tag in xml file.')
    if size['width'] is None:
        raise Exception('Could not find width tag in xml file.')
    if size['height'] is None:
        raise Exception('Could not find height tag in xml file.')
    # image_item = dict()    #按照一定的顺序，这里采用collections.OrderedDict()
    image_item = collections.OrderedDict()
    jpg_name = os.path.splitext(file_name)[0] + img_suffixes
    image_item['file_name'] = jpg_name
    image_item['width'] = size['width']
    image_item['height'] = size['height']
    # image_item['id'] = image_id
    # coco['images'].append(image_item)
    if is_train:
        image_item['id'] = train_image_id
        coco_train['images'].append(image_item)
        image_id = train_image_id
        train_image_id += 1
    else:
        image_item['id'] = val_image_id
        coco_val['images'].append(image_item)
        image_id = val_image_id
        val_image_id += 1
    image_set.add(jpg_name)
    image_id = image_id + 1
    return image_id


def addAnnoItem(object_name, image_id, category_id, bbox, is_train):
    global annotation_id
    # annotation_item = dict()
    annotation_item = collections.OrderedDict()
    annotation_item['segmentation'] = []
    seg = []
    # bbox[] is x,y,w,h
    # left_top
    seg.append(bbox[0])
    seg.append(bbox[1])
    # left_bottom
    seg.append(bbox[0])
    seg.append(bbox[1] + bbox[3])
    # right_bottom
    seg.append(bbox[0] + bbox[2])
    seg.append(bbox[1] + bbox[3])
    # right_top
    seg.append(bbox[0] + bbox[2])
    seg.append(bbox[1])
    annotation_item['segmentation'].append(seg)
    annotation_item['area'] = bbox[2] * bbox[3]
    annotation_item['iscrowd'] = 0
    annotation_item['image_id'] = image_id
    annotation_item['bbox'] = bbox
    annotation_item['category_id'] = category_id
    annotation_item['id'] = annotation_id
    annotation_item['ignore'] = 0
    annotation_id += 1
    # coco['annotations'].append(annotation_item)
    if is_train:
        coco_train['annotations'].append(annotation_item)
    else:
        coco_val['annotations'].append(annotation_item)

def parseXmlFiles(xml_path, xmllist, img_suffixes, is_train):
    for f in xmllist:
        if not f.endswith('.xml'):
            continue

        bndbox = dict()
        size = dict()
        current_image_id = None
        current_category_id = None
        file_name = None
        size['width'] = None
        size['height'] = None
        size['depth'] = None

        xml_file = os.path.join(xml_path, f)
        print(xml_file)

        tree = ET.parse(xml_file)
        root = tree.getroot()  # 抓根结点元素

        if root.tag != 'annotation':  # 根节点标签
            raise Exception('pascal voc xml root element should be annotation, rather than {}'.format(root.tag))

        # elem is , , ,