Yolo alexeyAB

Gaussian
https://github.com/AlexeyAB/darknet

https://github.com/jwchoi384/Gaussian_YOLOv3 Gaussian YOLOv3: An Accurate and Fast Object Detector Using Localization Uncertainty for Autonomous Driving and incorporated into Yolo alexeyAB, the preferred implementation. http://openaccess.thecvf.com/content_ICCV_2019/html/Choi_Gaussian_YOLOv3_An_Accurate_and_Fast_Object_Detector_Using_Localization_ICCV_2019_paper.htmlhe use of object detection algorithms is becoming increasingly important in autonomous vehicles, and object detection at high accuracy and a fast inference speed is essential for safe autonomous driving. A false positive (FP) from a false localization during autonomous driving can lead to fatal accidents and hinder safe and efficient driving. Therefore, a detection algorithm that can cope with mislocalizations is required in autonomous driving applications. This paper proposes a method for improving the detection accuracy while supporting a real-time operation by modeling the bounding box (bbox) of YOLOv3, which is the most representative of one-stage detectors, with a Gaussian parameter and redesigning the loss function. In addition, this paper proposes a method for predicting the localization uncertainty that indicates the reliability of bbox. By using the predicted localization uncertainty during the detection process, the proposed schemes can significantly reduce the FP and increase the true positive (TP), thereby improving the accuracy. Compared to a conventional YOLOv3, the proposed algorithm, Gaussian YOLOv3, improves the mean average precision (mAP) by 3.09 and 3.5 on the KITTI and Berkeley deep drive (BDD) datasets, respectively. Nevertheless, the proposed algorithm is capable of real-time detection at faster than 42 frames per second (fps) and shows a higher accuracy than previous approaches with a similar fps. Therefore, the proposed algorithm is the most suitable for autonomous driving applications.

AlexeyAB (preferred fork)
https://groups.google.com/forum/#!msg/darknet/8qC4k_cWgOc/TDxjY34ZBQAJ To save detection results into the txt file of objects from images, you can compile with LIBSO=1 and then do: ./uselib air.txt > result.txt Where air.txt contains paths to images. So result.txt will contain detection coords. To save detection results of objects from videofile, you can compile with LIBSO=1 and then do: ./uselib test.mp4 > result.txt before this, uncomment this line and recompile - but this slightly reduces FPS: https://github.com/AlexeyAB/darknet/blob/548a0bc652b562723695cc107f0844f11d1a2207/src/yolo_console_dll.cpp#L169 Also read: https://github.com/AlexeyAB/darknet/issues/125#issuecomment-320373088

пятница, 29 сентября 2017 г., 12:05:26 UTC+3 пользователь alex.ange...@gmail.com написал: https://github.com/pjreddie/darknet/issues/723 Run YoloV3 detections on thousands of images and save outputs?

./darknet detector test ./cfg/coco.data ./cfg/yolov3.cfg ./yolov3.weights -dont_show < data/train.txt > result.txt

https://github.com/AlexeyAB/darknet#how-to-train-pascal-voc-data Fork of Yolo, download android webcam app. use android phone as network camera input stream.
 * https://github.com/AlexeyAB/Yolo_mark GUI for marking bounded boxes of objects in images for training Yolo v2
 * Multi gpu training instructions.
 * https://github.com/pjreddie/darknet/pull/861 read directly from memory.
 * https://github.com/AlexeyAB/darknet/issues/407 Cuda patch.
 * https://timebutt.github.io/static/how-to-train-yolov2-to-detect-custom-objects/ sites https://pjreddie.com/darknet/yolo/ training data set.  from Nils Tijtgat. YOLOv2 is known to struggle when detecting small objects. The Darknet Google Groups has many different topics on how you could improve performance, you could have a look there to find inspiration. A suggestion that is often repeated is to train YOLOv2 using a higher input resolution, instead of 416x416. See this or this for instance. |sort:relevance/darknet/MumMJ2D8H9Y/n6nAIM0EAgAJ yolo small google groups1, yolo small 2 google groups
 * https://timebutt.github.io/static/understanding-yolov2-training-output/
 * |sort:relevance/darknet/BcsBQ-rez9Q/_BlMlIUSAQAJ sept2017 version is 15fps on tx1, use alexeyab as it gives 200fps. Also pjreddie reorganized his code substantially changing folders.
 * counting number of objects in image, people tracking, https://www.youtube.com/watch?v=QeWl0h3kQ24
 * fix for 4k video fps from 7 to 20
 * run demo without screen when using AMZ gpu's for example.
 * resize_image
 * small object detection
 * https://groups.google.com/forum/#!topic/darknet/EjQGffa7y-k  multi gpu training. See Nvidia cuda blade install for cuda install on linux..
 * Yolo bounding box

block classes
https://groups.google.com/forum/#!topic/darknet/mougiWkA1DI

Currently you can use this repository: https://github.com/AlexeyAB/darknet Use yolov3.cfg with https://pjreddie.com/media/files/yolov3.weights And just add dont_show message before each class that you don't want to detect, in this file: https://github.com/AlexeyAB/darknet/blob/master/cfg/coco.names person dont_show bicycle dont_show car dont_show motorbike

links
Nvidia Jetson