一、DetectNet

之前使用的模型是ImageNet跟GoogleNet的resNet-18，都是針對整張圖片或影片去做辨識，如果要辨識動態或一張照片上多個物體，我們就先來體驗detectNet吧，detectNet可以在一個frame偵測許多物件，很期待吧。相同的Jetson nano也提供了兩種版本的程式碼，分別是python跟c++。

以下是官方說明，其實說的蠻清楚的，高達91種分類別可以被辨識出來。

The detectNet object accepts an image as input, and outputs a list of coordinates of the detected bounding boxes along with their classes and confidence values. detectNet is available to use from Python and C++. See below for various pre-trained detection models available for download. The default model used is a 91-class SSD-Mobilenet-v2 model trained on the MS COCO dataset, which achieves realtime inferencing performance on Jetson with TensorRT. 

As examples of using the detectNet class, we provide sample programs for C++ and Python:

• detectnet.cpp (C++)

• detectnet.py (Python)

從偵測影像開始吧

我們先進入Docker container吧

$ cd jetson-inference

$ ./docker/run.sh

$  cd build/aarch64/bin/

可以一次辨識多個物件

二、一次辨識多個檔案

The 91-class MS COCO dataset that the SSD-based models were trained on include people, vehicles, animals, and assorted types of household objects to detect. 

看來可以多玩玩看其他網路下載的照片

1. 利用雙引號把images資料夾內所有peds_開頭的jpg檔進行辨識

2. 輸出到images/test資料夾，輸出的檔案為peds_output_流水號.jpg檔，編號從0開始

三、來試試看多人的吧

四、測試影片

上網下載影片 https://www.videezy.com/free-video/pedestrian

$ ./detectnet.py images/city.mp4 /images/test/city_out.mp4

影片太快的話...辨識還是沒那麼精準，而且Jetson nano也會有點lag，可能需要升級了

五、Running Different Detection Models

• 使用ssd-inception-v2 模組

• 切換到以下資料夾，跟著執行安裝 ssd-inception-v2 模組

$ cd jetson-inference/tools

$ ./download-models.sh

選擇 SSD-Inception-v2 ，然後Press Enter

一樣，第一次執行要跑很久喔

六、程式碼

先腦補python程式 import sys

跟之前的自己寫程式不一樣的有多了sys模組，sys.argv 是用來取得執行 Python 文件時命令列參數的方法 。sys.argv 其實就是個 list，除了sys.argv[0]以外，sys.argv 裡面存放著執行程式時帶入的外部參數 

可以參考 https://shengyu7697.github.io/blog/2019/12/28/Python-sys-argv/

再腦補 if __name__ == '__main__' -----這一段與下面程式無關啦...哈哈

就想放在這裡

if __name__ == '__main__' 涵義 這篇寫得很好 http://blog.castman.net/%E6%95%99%E5%AD%B8/2018/01/27/python-name-main.html

我把這一篇的排版重新順一下，超好理解

import jetson.inference

import jetson.utils

import argparse

import sys

# parse the command line

parser = argparse.ArgumentParser(description="Locate objects in a live camera stream using an object detection DNN.")

parser.add_argument("input_URI", type=str, default="", nargs='?', help="URI of the input stream")

parser.add_argument("output_URI", type=str, default="", nargs='?', help="URI of the output stream")

parser.add_argument("--network", type=str, default="ssd-mobilenet-v2", help="pre-trained model to load (see below for options)")

parser.add_argument("--overlay", type=str, default="box,labels,conf", help="detection overlay flags (e.g. --overlay=box,labels,conf)\nvalid combinations are:  'box', 'labels', 'conf', 'none'")

parser.add_argument("--threshold", type=float, default=0.5, help="minimum detection threshold to use") 

try:

opt = parser.parse_known_args()[0]

except:

print("")

parser.print_help()

sys.exit(0)

# load the object detection network

net = jetson.inference.detectNet(opt.network, sys.argv, opt.threshold)

# create video sources & outputs

input = jetson.utils.videoSource(opt.input_URI, argv=sys.argv)

output = jetson.utils.videoOutput(opt.output_URI, argv=sys.argv)

# process frames until the user exits

while True:

# capture the next image

img = input.Capture()

# detect objects in the image (with overlay)

detections = net.Detect(img, overlay=opt.overlay)

# print the detections

print("detected {:d} objects in image".format(len(detections)))

for detection in detections:

print(detection)

# render the image

output.Render(img)

# update the title bar

output.SetStatus("{:s} | Network {:.0f} FPS".format(opt.network, net.GetNetworkFPS()))

# print out performance info

net.PrintProfilerTimes()

# exit on input/output EOS

if not input.IsStreaming() or not output.IsStreaming():

break