对比最终实验报告--罗宇鑫

:::info 该报告汇报时间✏️： 2021-11-6

汇报内容涉及模型：

• Faster - vgg
• YoloV4
• SSD

大礼包下载地址：https://yun.hengyimonster.top/s/yrc6
下载密码：iww527
有效期：2021年11月6日即日起30天有效期
解压密码：zhebuchongjidinghui :::

概述：

HfsCk]!pQC-H\5BAc^gBo/^qq)/i21ufiN@H”Y

uR%{)Y’Qz-n3oGU`ZJo@(1ntxp8U1+bW;JlZH^I4%0rxf;[N+eQ)Lolrw&E%,4q1

1. SSD - Pytorch

   仓库地址： https://github.com/bubbliiiing/ssd-pytorch 预训练权重地址： https://github.com/bubbliiiing/ssd-pytorch/releases/download/v1.0/ssd_weights.pth

2. yolov4-Pytorch

   仓库地址：https://github.com/bubbliiiing/yolov4-pytorch 预训练权重地址：https://github.com/bubbliiiing/yolov4-pytorch/releases/download/v1.0/yolo4_voc_weights.pth

3. fater-rcnn-pytorch（vgg版本）

   仓库地址：https://github.com/bubbliiiing/faster-rcnn-pytorch 预训练权重：https://github.com/bubbliiiing/faster-rcnn-pytorch/releases/download/v1.0/voc_weights_vgg.pth

4. faster-rcnn.pytorch

   仓库地址：https://github.com/jwyang/faster-rcnn.pytorch 预训练权重：https://filebox.ece.vt.edu/~jw2yang/faster-rcnn/pretrained-base-models/vgg16_caffe.pth

Flops：

SSD - Pytorch

Flops：68.23 GMac
计算代码：

from nets.ssd import SSD300
import torchvision.models as models
from ptflops import get_model_complexity_info
def get_classes(classes_path): # 去取name和数量
    with open(classes_path, encoding='utf-8') as f:
        class_names = f.readlines()
    class_names = [c.strip() for c in class_names]
    return class_names, len(class_names)
if __name__ == "__main__":
    classes_path = 'model_data/voc_classes.txt'
    class_names, num_classes  = get_classes(classes_path)
    # 对应的模型创建
    myNet = SSD300(num_classes + 1, 'vgg')  
    # 根据层数 以及跑的图片大小进行设置
    flops, params = get_model_complexity_info(myNet, (3,416,416), as_strings=True, print_per_layer_stat=True)
    print("Flops: {}".format(flops))
    print("Params: " + params)

FLOPs-yolov4

Flops：30.28 GMac

from nets.yolo import YoloBody
from ptflops import get_model_complexity_info
def get_classes(classes_path): # 去取name和数量
    with open(classes_path, encoding='utf-8') as f:
        class_names = f.readlines()
    class_names = [c.strip() for c in class_names]
    return class_names, len(class_names)
if __name__ == "__main__":
    classes_path = 'model_data/voc_clothes_classes.txt'
    class_names, num_classes  = get_classes(classes_path)
    # 对应的模型创建
    anchors_mask = [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
    myNet = YoloBody(anchors_mask, num_classes)  
    # 根据层数 以及跑的图片大小进行设置
    flops, params = get_model_complexity_info(myNet, (3,416,416), as_strings=True, print_per_layer_stat=True)
    print("Flops: {}".format(flops))
    print("Params: " + params)

Faster-rcnn

Flops：91.22 GMac

from nets.frcnn import FasterRCNN
import torchvision.models as models
from ptflops import get_model_complexity_info
def get_classes(classes_path):
    with open(classes_path, encoding='utf-8') as f:
        class_names = f.readlines()
    class_names = [c.strip() for c in class_names]
    return class_names, len(class_names)
if __name__ == "__main__":
    # print("Load model.")
    # ssd = SSD(confidence = 0.01, nms_iou = 0.5)
    # print("Load model done.")
    # net = models.vgg16() #可以为自己搭建的模型
    classes_path = 'model_data/voc_classes.txt'
    class_names, num_classes  = get_classes(classes_path)
    myNet = FasterRCNN(num_classes, "predict", anchor_scales = [8, 16, 32], backbone = 'vgg')
    flops, params = get_model_complexity_info(myNet, (3,416,416), as_strings=True, print_per_layer_stat=True)
    print("Flops: {}".format(flops))
    print("Params: " + params)

实验记录

SSD - Pytorch（老标签）

YoloV4（老标签）

Faster-rcnn（老标签）

YoloV4（新标签）

SSD-Pytorch（新标签）

Faster-rcnn(新标签)

Faster-rcnn（新库）

注意⚠️： 没有生成对应的文件图，请到对应路径下跑～

SSD-Pytorch(排序标签)

YoloV4-Pytroch（排序标签）

饭后甜点

SSD

YoloV4

Faster-rcnn