工業級別的目標檢測關注的不僅僅是精度,還有速度,能達到實時是最理想狀態,一般來講,目標檢測實時大于12.5fps被認為是實時,針對TX2利用yolov4檢測博主做了一個詳細的調研和測驗,
1.下載darknet,網址如下:
git clone https://github.com/AlexeyAB/darknet.git
2.配置makefile檔案
由于TX2已經刷機Jetpack4.4,TX2里面有gpu,cuda和cudnn等,修改makefile檔案如下:
GPU=1
CUDNN=1
OPENCV=1

3.在darknet路徑下編譯如下
make
4.下載權重,放到darknet目錄下
# yolov4-tiny.weights
wget https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.weights
# yolov4.weights
wget https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4.weights
5.測驗(yolov4 and yolov4-tiny)
(1).測驗圖片
./darknet detector test cfg/coco.data cfg/yolov4-tiny.cfg yolov4-tiny.weights data/dog.jpg
./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights data/dog.jpg
(2).測驗視頻
./darknet detector demo cfg/coco.data cfg/yolov4-tiny.cfg yolov4-tiny.weights data/sample.mp4
./darknet detector demo cfg/coco.data cfg/yolov4-tiny.cfg yolov4-tiny.weights data/sample.mp4
(3).實時測驗板載攝像頭 (CSI攝像頭實時檢測)
./darknet detector demo cfg/coco.data cfg/yolov4-tiny.cfg yolov4-tiny.weights "nvarguscamerasrc ! video/x-raw(memory:NVMM), width=1280, height=720, format=NV12, framerate=30/1 ! nvvidconv flip-method=0 ! video/x-raw, width=1280, height=720, format=BGRx ! videoconvert ! video/x-raw, format=BGR ! appsink"
./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights "nvarguscamerasrc ! video/x-raw(memory:NVMM), width=1280, height=720, format=NV12, framerate=30/1 ! nvvidconv flip-method=0 ! video/x-raw, width=1280, height=720, format=BGRx ! videoconvert ! video/x-raw, format=BGR ! appsink"
(4).實時檢測usb攝像頭
./darknet detector demo cfg/coco.data cfg/yolov4-tiny.cfg yolov4-tiny.weights -c 1
./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights -c 1

(5).rstp實時檢測
./darknet detector demo cfg/coco.data cfg/yolov4-tiny.cfg yolov4-tiny.weights rstp://admin:admin@20.10.7.34/0
./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights rstp://admin:admin@20.10.7.34/0
6 總結
評論區留言哦
下集預告:如何按照自己的需求訓練模型以及二次開發
轉載請註明出處,本文鏈接:https://www.uj5u.com/qita/242467.html
標籤:其他
