fairmot代码解析（自己在看代码做笔记用）

本文详细解析了FairMOT框架中的关键代码部分，包括视频加载、模型推理（网络前向传播、目标检测与关联）、模型定义（如DLASeg网络结构）、非极大值抑制算法、以及目标跟踪器的管理机制。此外，还涉及模型训练过程中的heads设置和模型加载。

fairmot代码解析

1.加载视频
2.模型推理
3.训练
- 3.1输出的heads确定
- 3.2加载模型

1.加载视频

dataloader 是在jde.py函数中定义，用于加载视频，以及进行图像增强。

import datasets.dataset.jde as datasets
dataloader = datasets.LoadVideo(opt.input_video, opt.img_size)

其中LoadVideo如下

class LoadVideo:  # for inference
    def __init__(self, path, img_size=(1088, 608)):
        self.cap = cv2.VideoCapture(path)
        self.frame_rate = int(round(self.cap.get(cv2.CAP_PROP_FPS)))
        self.vw = int(self.cap.get(cv2.CAP_PROP_FRAME_WIDTH))
        self.vh = int(self.cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
        self.vn = int(self.cap.get(cv2.CAP_PROP_FRAME_COUNT))

        self.width = img_size[0]
        self.height = img_size[1]
        self.count = 0

        self.w, self.h = 1920, 1080
        print('Lenth of the video: {:d} frames'.format(self.vn))

    def get_size(self, vw, vh, dw, dh):
        wa, ha = float(dw) / vw, float(dh) / vh
        a = min(wa, ha)
        return int(vw * a), int(vh * a)

    def __iter__(self):
        self.count = -1
        return self

    def __next__(self):
        self.count += 1
        if self.count == len(self):
            raise StopIteration
        # Read image
        res, img0 = self.cap.read()  # BGR
        assert img0 is not None, 'Failed to load frame {:d}'.format(self.count)
        img0 = cv2.resize(img0, (self.w, self.h))

        # Padded resize
        img, _, _, _ = letterbox(img0, height=self.height, width=self.width)

        # Normalize RGB
        img = img[:, :, ::-1].transpose(2, 0, 1)
        img = np.ascontiguousarray(img, dtype=np.float32)
        img /= 255.0

        # cv2.imwrite(img_path + '.letterbox.jpg', 255 * img.transpose((1, 2, 0))[:, :, ::-1])  # save letterbox image
        return self.count, img, img0

    def __len__(self):
        return self.vn  # number of files