What's the different between Darknet here and the official one?

[XP-CHRC]

Hi,everyone, I run the same yolov2-tiny.cfg and yolov2-tiny.weights in the same device(jetson tx2), but the darknet download here is faster than the official one. So what's the difference between them?
Thanks.

[AlexeyAB]

Hi,

Did you use ./darknet detector demo... command?

There are many changes such as:

• added support for Windows
• improved binary neural network performance 2x-4x times for Detection on CPU and GPU if you trained your own weights by using this XNOR-net model (bit-1 inference) : https://github.com/AlexeyAB/darknet/blob/master/cfg/yolov3-tiny_xnor.cfg
• improved neural network performance ~7% by fusing 2 layers into 1: Convolutional + Batch-norm
• improved neural network performance Detection 3x times, Training 2 x times on GPU Volta (Tesla V100, Titan V, ...) using Tensor Cores if CUDNN_HALF defined in the Makefile or darknet.sln
• improved performance ~1.2x times on FullHD, ~2x times on 4K, for detection on the video (file/stream) using darknet detector demo...
• improved performance 3.5 X times of data augmentation for training (using OpenCV SSE/AVX functions instead of hand-written functions) - removes bottleneck for training on multi-GPU or GPU Volta
• improved performance of detection and training on Intel CPU with AVX (Yolo v3 ~85%, Yolo v2 ~10%)
• fixed usage of [reorg]-layer
• optimized memory allocation during network resizing when random=1
• optimized initialization GPU for detection - we use batch=1 initially instead of re-init with batch=1
• added correct calculation of mAP, F1, IoU, Precision-Recall using command darknet detector map...
• added drawing of chart of average loss during training
• added calculation of anchors for training
• added example of Detection and Tracking objects: https://github.com/AlexeyAB/darknet/blob/master/src/yolo_console_dll.cpp
• fixed code for use Web-cam on OpenCV 3.x
• run-time tips and warnings if you use incorrect cfg-file or dataset
• many other fixes of code...

And added manual How to train yolo.

[RushNuts]

How much fps on tx2?

[XP-CHRC]

@AlexeyAB yes, i used ./darknet detector demo... command, and i want to know what make the demo faster here than the official one? Did you implement some optimization tricks?
thanks for your reply，you are so 牛逼！

[XP-CHRC]

@RushNuts 416*416 About 15fps, less than 20, without any optimizations. and what's more the speed depend on the quality of the video , i don't know why?

[AlexeyAB]

@XP-CHRC

15 fps on TX2 using this fork?
And how much fps on TX2 using original fork for yolov2-tiny.cfg?

There is optimization in resizing video frame to the network size, I just use OpenCV implementation cvResize() instead of hand-written implementation - it was a bottleneck - perfomance depends on initial resolution of video:

darknet/src/demo.c

Lines 64 to 65 in d0039f6

	//in = get_image_from_stream(cap);
	in = get_image_from_stream_resize(cap, net.w, net.h, &in_img, use_webcam);

darknet/src/image.c

Line 874 in d0039f6

cvResize(src, new_img, CV_INTER_LINEAR);

[gangyahaidao]

this darknet is faster, it's true! Then i am confused, Why don't official projects adopt this project's advantages?
A Great Job, Thanks @AlexeyAB

[XP-CHRC]

@AlexeyAB Input size: 416*416, about 10fps on tx2 when use the original fork for yolov2-tiny.cfg.
thanks for your jobs and reply.

[RushNuts]

@XP-CHRC Oh, 15fps. Not bad. What's resolution your input video?

[XP-CHRC]

@RushNuts 416*416