search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.63k stars 7.95k forks source link

YOLOv4-tiny released: 40.2% AP50, 371 FPS (GTX 1080 Ti), 1770 FPS tkDNN/TensorRT #6067

Open AlexeyAB opened 4 years ago

AlexeyAB commented 4 years ago

Discussion: https://www.reddit.com/r/MachineLearning/comments/hu7lyt/p_yolov4tiny_speed_1770_fps_tensorrtbatch4/

Full structure: structure of yolov4-tiny.cfg model

YOLOv4-tiny released: 40.2% AP50, 371 FPS (GTX 1080 Ti) / 330 FPS (RTX 2070)

cmp

source yolov3-tiny (800x800) yolov4-tiny (800x800)
src predictions predictions
laclouis5 commented 4 years ago

@AlexeyAB Really great work!

For training, same partial weights as old tiny Yolo i.e. the first 15 layers yolov4-tiny.conv.15?

AlexeyAB commented 4 years ago

@laclouis5

Use this pre-trained file for trainint yolov4-tiny.cfg: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.conv.29

How to train yolov4-tiny.cfg: https://github.com/AlexeyAB/darknet#how-to-train-tiny-yolo-to-detect-your-custom-objects

muhammad-maaz-confiz commented 4 years ago

Hi @alexeab,

Does OpenCV DNN module supports YoloV4-tiny? Thanks

AlexeyAB commented 4 years ago

We are waiting for the implementation of the YOLOv4-tiny in libraries:

ark- commented 4 years ago

Hi @alexeab,

Does OpenCV DNN module supports YoloV4-tiny? Thanks

OpenCV implemented it in their master branch in 6 days for Yolov4. This looks like a more trivial change required so here's hoping it will be live in a few days.

LukeAI commented 4 years ago

@AlexeyAB v. exciting. are you planning to release a paper on it? Would love to read some details about how it works. Is it a novel backbone or one of the existing CSPs? Do you have any numbers on the performance of the backbone as a classifier?

AlexeyAB commented 4 years ago

@LukeAI

Bao150297 commented 4 years ago

@AlexeyAB Can you explain why the last yolo layer uses masks starting from 1, not 0?

CSTEZCAN commented 4 years ago

confirmed. performance is WOW. https://youtu.be/TWteusBINIw offline test (without connecting to the stream) Screenshot from 2020-06-26 14-57-03

AlexeyAB commented 4 years ago

@CSTEZCAN Hi, What FPS can you get by using such command without mjpeg_port? darknet.exe detector demo cfg/coco.data cfg/yolov4-tiny.cfg yolov4-tiny.weights -i 0 -thresh 0.25 -ext_output test.mp4 -dont_show

On GPU RTX 2070, CPU Core i7 6700K

AlexeyAB commented 4 years ago

@DoriHp Just to compare with Yolov3-tiny where were used the same masks, it seems tiny models don't detect well small objects anyway.

Bao150297 commented 4 years ago

I saw yolov3-tiny_3l.cfg with 3 yolo layers. So due to what you said, the last yolo layers has no use?

AlexeyAB commented 4 years ago

To detect small objects you must also use 3-yolo-layers in yolov4-tiny.

CSTEZCAN commented 4 years ago

@AlexeyAB Hello Alexey! this is your "Piano Concerto No. 2 Op. 18". I have infinite respect for your work.

My results as follows;

nvidia-smi -pl 125 watts titan x pascal Screenshot from 2020-06-26 20-15-18

nvidia-smi -pl 250 watts titan x pascal Screenshot from 2020-06-26 20-16-03

Will be running tests on Jetson Nano, TX2 and Xavier later..

AlexeyAB commented 4 years ago

@CSTEZCAN Thanks! I think yolov4-tiny can work with 500 - 1000 FPS by using OpenCV or tkDNN/TensorRT when it will be implemented in these libraries.

CSTEZCAN commented 4 years ago

@AlexeyAB I have no doubt. The only weird thing I noticed is, it uses CPU relatively more during training compared to YOLOv4. The recommended system must start from Ryzen 3500 and above for an optimal performance (if you are creating such recommended setup list) :)

choochtech commented 4 years ago

@AlexeyAB good job ! great work !

CSTEZCAN commented 4 years ago

@AlexeyAB Nvidia Jetson AGX Xavier can do (all avg_fps) MAX_N : 120.5 15W: 36.0 30W ALL: 68.9 30W 2core: 28.9

Screenshot from 2020-06-27 01-38-20

Training on tiny takes around 2 hours for a single class model for 5000 batches (which is usually enough) on xavier

it is just great!

ghost commented 4 years ago

@AlexeyAB you truly outdone yourself, I am impressed. The OpenCV/tensort rt will be a game changer. @CSTEZCAN post the nano results, along with testing cpu-only mode results.

marvision-ai commented 4 years ago

@CSTEZCAN what jetpack version are you using ? And are you training networks on the Xavier too?

1027663760 commented 4 years ago

On training on yolov4-tiny, I met nan, and when adversarial_lr=1 attention=1, I saw a pure black picture The default learning_rate is too large, I set it to learning_rate=0.001 to only ease the time when nan appears

AlexeyAB commented 4 years ago

@AlexeyAB you truly outdone yourself, I am impressed. The OpenCV/tensort rt will be a game changer.

Thanks so much for @WongKinYiu

CSTEZCAN commented 4 years ago

@deepseek nano has some problems, maybe about ram. edit2: nano can do max_N:16fps, 5W:10.2fps. @mbufi I've trained 1 model on it, at 30 watts mode, it can train models without any problems. edit1: it is an old jetpack from last year, probably 4.2 or 4.3.

CSTEZCAN commented 4 years ago

There is a small question - there is no implementation in cv2 yet - what is the easiest way to run the model on the CPU (in the application) without a large number of dependencies?

question of the century :)

LukeAI commented 4 years ago

There is a small question - there is no implementation in cv2 yet - what is the easiest way to run the model on the CPU (in the application) without a large number of dependencies?

can you not run with darknet, compiled without CUDA?

marvision-ai commented 4 years ago

@CSTEZCAN what resolution images are you training and running inference on with the Xavier? I'd love to see an example cfg file that supports that performance you are showing.

Thanks for testing everything! Super helpful.

CSTEZCAN commented 4 years ago

@CSTEZCAN what resolution images are you training and running inference on with the Xavier? I'd love to see an example cfg file that supports that performance you are showing.

nothing especially changed on the yolov4-tiny.cfg (except class number and filters) so training & inferencing resolutions are same. https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-tiny.cfg

zbyuan commented 4 years ago

@AlexeyAB Hello Alexey! [route] layers=-1 groups=2 group_id=1 What does it mean？

AlexeyAB commented 4 years ago

@zbyuan Read: https://github.com/opencv/opencv/issues/17666

CSTEZCAN commented 4 years ago

@CSTEZCAN okay cool. And for the video you are passing it, what's the resolution of the frames before resizing them? How are the detections of small objects? (especially if the algorithm is resizing it to smaller before putting it through the network.) This is great information you are providing.

Dude come on, I don't know your data, I don't know your label quality. You have to test them "yourself"

marvision-ai commented 4 years ago

@CSTEZCAN great thanks!

CSTEZCAN commented 4 years ago

mac mini perf test. https://www.youtube.com/watch?v=0ZrfTglY4SI

wwzh2015 commented 4 years ago

Hi，What is the score that the yolo4-tiny tested in voc2007 and trained in voc2007+2012.

piaomiaoju commented 4 years ago

why there are no spp layers selected in yolo4-tiny? @AlexeyAB

AlexeyAB commented 4 years ago

@piaomiaoju @WongKinYiu Yes, maybe we can try to use small SPP: yolov4-tiny-spp.cfg.txt

### SPP

[maxpool]
size=3
stride=1

[route]
layers=-2

[maxpool]
size=5
stride=1

[route]
layers=-1,-3,-4
Anafeyka commented 4 years ago

@piaomiaoju @WongKinYiu Yes, maybe we can try to use small SPP: yolov4-tiny-spp.cfg.txt

### SPP

[maxpool]
size=3
stride=1

[route]
layers=-2

[maxpool]
size=5
stride=1

[route]
layers=-1,-3,-4

share an example cfg small SPP + 3 layers of yolo. To detect small objects

BackT0TheFuture commented 4 years ago

To detect small objects you must also use 3-yolo-layers in yolov4-tiny.

great works! do you have plan for 3-yolo-layers in yolov4-tiny? thanks!

javier-box commented 4 years ago

@AlexeyAB .. nice... well done.. yolov4-tiny is in the goldielocks zone... not too slow and just the right mAP.

You can also use the other yolov4 bag of tricks to get a slightly better mAP

www7890 commented 4 years ago

To detect small objects you must also use 3-yolo-layers in yolov4-tiny.

great works! do you have plan for 3-yolo-layers in yolov4-tiny? thanks!

Want to know , too!

tobyglei commented 4 years ago

Exciting news. Thanks for the good work @AlexeyAB

JasonDoingGreat commented 4 years ago

@CSTEZCAN Thanks! I think yolov4-tiny can work with 500 - 1000 FPS by using OpenCV or tkDNN/TensorRT when it will be implemented in these libraries.

Using tkDNN fresh yolov4-tiny impl, tested on Jetson Nano with JetPack 4.4, TensorRT v7.1, 416 input size

For FP32, profile results:

Time stats:
Min: 37.3371 ms
Max: 122.952 ms
Avg: 38.0922 ms 26.2521 FPS

For FP16, profile results:

Time stats:
Min: 24.5687 ms
Max: 90.5088 ms
Avg: 25.5292 ms 39.1709 FPS
piaomiaoju commented 4 years ago

[route] layers=-1 groups=2 group_id=1 if we don't slice the layer here, i think there are no accuracy degradation, but very little size added.

wwzh2015 commented 4 years ago

degrade

what‘s mean?

AlexeyAB commented 4 years ago

RTX 2080Ti (CUDA 10.2, TensorRT 7.0.0, Cudnn 7.6.5, tkDNN); for yolo4tiny 416x416, on 1200 images of size 416x416: https://github.com/ceccocats/tkDNN/issues/59#issuecomment-652269964

AlexeyAB commented 4 years ago

@CSTEZCAN Thanks! I think yolov4-tiny can work with 500 - 1000 FPS by using OpenCV or tkDNN/TensorRT when it will be implemented in these libraries.

I was very mistaken, here 2000 FPS, not 1000 FPS

AlexeyAB commented 4 years ago

Jetson AGX (CUDA 10.2, TensorRT 7.0.0, Cudnn 7.6.5, tkDNN); for yolo4tiny 416x416, on 1200 images of size 416x416: https://github.com/ceccocats/tkDNN/issues/59#issuecomment-652420971

ghost commented 4 years ago

Mayday, Mayday!! @CSTEZCAN Confirm the fps math, does it check? because Alex might have changed the GAME. @AlexeyAB How is this even possible??! were you able to main the accuracy? tell me everything.

CSTEZCAN commented 4 years ago

@CSTEZCAN Thanks! I think yolov4-tiny can work with 500 - 1000 FPS by using OpenCV or tkDNN/TensorRT when it will be implemented in these libraries.

I was very mistaken, here 2000 FPS, not 1000 FPS

I really don't know what to say @AlexeyAB you've rocked this planet & probably beyond :) Lets work on totally unsupervised classification & detection next time for mars rovers :)

CSTEZCAN commented 4 years ago

Mayday, Mayday!! @CSTEZCAN Confirm the fps math, does it check? because Alex might have changed the GAME. @AlexeyAB How is this even possible??!

I don't have tkDNN setup on Xavier and I don't have RTX GPU to check INT8, FP16 to reach those numbers... but the source is trustworthy. If they say it works, it works.

AlexeyAB commented 4 years ago

Darknet isn't optimized for inference on CPU (even with AVX=1 OPENMP=1). Use https://github.com/Tencent/ncnn for testing on CPU (Desktop, Laptop, Smartphones) or wait for the implementation of yolov4-tiny in the OpenCV library https://github.com/opencv/opencv/issues/17666