Perform a series of ablation experiments on yolov5 to make it lighter (smaller Flops, lower memory, and fewer parameters) and faster (add shuffle channel, yolov5 head for channel reduce. It can infer at least 10+ FPS On the Raspberry Pi 4B when input the frame with 320ร320) and is easier to deploy (removing the Focus layer and four slice operations, reducing the model quantization accuracy to an acceptable range).
ID | Model | Input_size | Flops | Params | Size๏ผM๏ผ | [email protected] | [email protected]:0.95 |
---|---|---|---|---|---|---|---|
001 | yolo-fastest | 320ร320 | 0.25G | 0.35M | 1.4 | 24.4 | - |
002 | YOLOv5-Liteeours | 320ร320 | 0.73G | 0.78M | 1.7 | 35.1 | - |
003 | NanoDet-m | 320ร320 | 0.72G | 0.95M | 1.8 | - | 20.6 |
004 | yolo-fastest-xl | 320ร320 | 0.72G | 0.92M | 3.5 | 34.3 | - |
005 | YOLOXNano | 416ร416 | 1.08G | 0.91M | 7.3(fp32) | - | 25.8 |
006 | yolov3-tiny | 416ร416 | 6.96G | 6.06M | 23.0 | 33.1 | 16.6 |
007 | yolov4-tiny | 416ร416 | 5.62G | 8.86M | 33.7 | 40.2 | 21.7 |
008 | YOLOv5-Litesours | 416ร416 | 1.66G | 1.64M | 3.4 | 42.0 | 25.2 |
009 | YOLOv5-Litecours | 512ร512 | 5.92G | 4.57M | 9.2 | 50.9 | 32.5 |
010 | NanoDet-EfficientLite2 | 512ร512 | 7.12G | 4.71M | 18.3 | - | 32.6 |
011 | YOLOv5s(6.0) | 640ร640 | 16.5G | 7.23M | 14.0 | 56.0 | 37.2 |
012 | YOLOv5-Litegours | 640ร640 | 15.6G | 5.39M | 10.9 | 57.6 | 39.1 |
See the wiki: https://github.com/ppogg/YOLOv5-Lite/wiki/Test-the-map-of-models-about-coco
Equipment | Computing backend | System | Input | Framework | v5lite-e | v5lite-s | v5lite-c | v5lite-g | YOLOv5s |
---|---|---|---|---|---|---|---|---|---|
Inter | @i5-10210U | window(x86) | 640ร640 | openvino | - | - | 46ms | - | 131ms |
Nvidia | @RTX 2080Ti | Linux(x86) | 640ร640 | torch | - | - | - | 15ms | 14ms |
Redmi K30 | @Snapdragon 730G | Android(armv8) | 320ร320 | ncnn | 27ms | 38ms | - | - | 163ms |
Xiaomi 10 | @Snapdragon 865 | Android(armv8) | 320ร320 | ncnn | 10ms | 14ms | - | - | 163ms |
Raspberrypi 4B | @ARM Cortex-A72 | Linux(arm64) | 320ร320 | ncnn | - | 84ms | - | - | 371ms |
Raspberrypi 4B | @ARM Cortex-A72 | Linux(arm64) | 320ร320 | mnn | - | 71ms | - | - | 356ms |
AXera-Pi | Cortex A7@CPU 3.6TOPs @NPU |
Linux(arm64) | 640ร640 | axpi | - | - | - | 22ms | 22ms |
https://zhuanlan.zhihu.com/p/672633849
- The above is a 4-thread test benchmark
- Raspberrypi 4B enable bf16s optimization๏ผRaspberrypi 64 Bit OS
ๅ ฅ็พค็ญๆก:ๅชๆ or ่ธ้ฆ or ้ๅ or ไฝ็งฉๅ่งฃ๏ผไปปๆๅ ถไธๅๅฏ๏ผ
Model | Size | Backbone | Head | Framework | Design for |
---|---|---|---|---|---|
v5Lite-e.pt | 1.7m | shufflenetv2๏ผMegvii๏ผ | v5Litee-head | Pytorch | Arm-cpu |
v5Lite-e.bin v5Lite-e.param |
1.7m | shufflenetv2 | v5Litee-head | ncnn | Arm-cpu |
v5Lite-e-int8.bin v5Lite-e-int8.param |
0.9m | shufflenetv2 | v5Litee-head | ncnn | Arm-cpu |
v5Lite-e-fp32.mnn | 3.0m | shufflenetv2 | v5Litee-head | mnn | Arm-cpu |
v5Lite-e-fp32.tnnmodel v5Lite-e-fp32.tnnproto |
2.9m | shufflenetv2 | v5Litee-head | tnn | arm-cpu |
v5Lite-e-320.onnx | 3.1m | shufflenetv2 | v5Litee-head | onnxruntime | x86-cpu |
Model | Size | Backbone | Head | Framework | Design for |
---|---|---|---|---|---|
v5Lite-s.pt | 3.4m | shufflenetv2๏ผMegvii๏ผ | v5Lites-head | Pytorch | Arm-cpu |
v5Lite-s.bin v5Lite-s.param |
3.3m | shufflenetv2 | v5Lites-head | ncnn | Arm-cpu |
v5Lite-s-int8.bin v5Lite-s-int8.param |
1.7m | shufflenetv2 | v5Lites-head | ncnn | Arm-cpu |
v5Lite-s.mnn | 3.3m | shufflenetv2 | v5Lites-head | mnn | Arm-cpu |
v5Lite-s-int4.mnn | 987k | shufflenetv2 | v5Lites-head | mnn | Arm-cpu |
v5Lite-s-fp16.bin v5Lite-s-fp16.xml |
3.4m | shufflenetv2 | v5Lites-head | openvivo | x86-cpu |
v5Lite-s-fp32.bin v5Lite-s-fp32.xml |
6.8m | shufflenetv2 | v5Lites-head | openvivo | x86-cpu |
v5Lite-s-fp16.tflite | 3.3m | shufflenetv2 | v5Lites-head | tflite | arm-cpu |
v5Lite-s-fp32.tflite | 6.7m | shufflenetv2 | v5Lites-head | tflite | arm-cpu |
v5Lite-s-int8.tflite | 1.8m | shufflenetv2 | v5Lites-head | tflite | arm-cpu |
v5Lite-s-416.onnx | 6.4m | shufflenetv2 | v5Lites-head | onnxruntime | x86-cpu |
Model | Size | Backbone | Head | Framework | Design for |
---|---|---|---|---|---|
v5Lite-c.pt | 9m | PPLcnet๏ผBaidu๏ผ | v5s-head | Pytorch | x86-cpu / x86-vpu |
v5Lite-c.bin v5Lite-c.xml |
8.7m | PPLcnet | v5s-head | openvivo | x86-cpu / x86-vpu |
v5Lite-c-512.onnx | 18m | PPLcnet | v5s-head | onnxruntime | x86-cpu |
Model | Size | Backbone | Head | Framework | Design for |
---|---|---|---|---|---|
v5Lite-g.pt | 10.9m | Repvgg๏ผTsinghua๏ผ | v5Liteg-head | Pytorch | x86-gpu / arm-gpu / arm-npu |
v5Lite-g-int8.engine | 8.5m | Repvgg-yolov5 | v5Liteg-head | Tensorrt | x86-gpu / arm-gpu / arm-npu |
v5lite-g-int8.tmfile | 8.7m | Repvgg-yolov5 | v5Liteg-head | Tengine | arm-npu |
v5Lite-g-640.onnx | 21m | Repvgg-yolov5 | yolov5-head | onnxruntime | x86-cpu |
v5Lite-g-640.joint | 7.1m | Repvgg-yolov5 | yolov5-head | axpi | arm-npu |
v5lite-e.pt
: | Baidu Drive | Google Drive ||โโโโโโ
ncnn-fp16
: | Baidu Drive | Google Drive |
|โโโโโโncnn-int8
: | Baidu Drive | Google Drive |
|โโโโโโmnn-e_bf16
: | Google Drive |
|โโโโโโmnn-d_bf16
: | Google Drive|
โโโโโโโonnx-fp32
: | Baidu Drive | Google Drive |
v5lite-s.pt
: | Baidu Drive | Google Drive ||โโโโโโ
ncnn-fp16
: | Baidu Drive | Google Drive |
|โโโโโโncnn-int8
: | Baidu Drive | Google Drive |
โโโโโโโtengine-fp32
: | Baidu Drive | Google Drive |
v5lite-c.pt
: Baidu Drive | Google Drive |โโโโโโโ
openvino-fp16
: | Baidu Drive | Google Drive |
v5lite-g.pt
: | Baidu Drive | Google Drive |โโโโโโโ
axpi-int8
: Google Drive |
Baidu Drive Password: pogg
v5lite-s model: TFLite Float32, Float16, INT8, Dynamic range quantization, ONNX, TFJS, TensorRT, OpenVINO IR FP32/FP16, Myriad Inference Engin Blob, CoreML
https://github.com/PINTO0309/PINTO_model_zoo/tree/main/180_YOLOv5-Lite
Thanks for PINTO0309:https://github.com/PINTO0309/PINTO_model_zoo/tree/main/180_YOLOv5-Lite
Install
Python>=3.6.0 is required with all requirements.txt installed including PyTorch>=1.7:
$ git clone https://github.com/ppogg/YOLOv5-Lite
$ cd YOLOv5-Lite
$ pip install -r requirements.txt
Inference with detect.py
detect.py
runs inference on a variety of sources, downloading models automatically from
the latest YOLOv5-Lite release and saving results to runs/detect
.
$ python detect.py --source 0 # webcam
file.jpg # image
file.mp4 # video
path/ # directory
path/*.jpg # glob
'https://youtu.be/NUsoVlDFqZg' # YouTube
'rtsp://example.com/media.mp4' # RTSP, RTMP, HTTP stream
Training
$ python train.py --data coco.yaml --cfg v5lite-e.yaml --weights v5lite-e.pt --batch-size 128
v5lite-s.yaml v5lite-s.pt 128
v5lite-c.yaml v5lite-c.pt 96
v5lite-g.yaml v5lite-g.pt 64
If you use multi-gpu. It's faster several times:
$ python -m torch.distributed.launch --nproc_per_node 2 train.py
DataSet
Training set and test set distribution ๏ผthe path with xx.jpg๏ผ
train: ../coco/images/train2017/
val: ../coco/images/val2017/
โโโ images # xx.jpg example
โ โโโ train2017
โ โ โโโ 000001.jpg
โ โ โโโ 000002.jpg
โ โ โโโ 000003.jpg
โ โโโ val2017
โ โโโ 100001.jpg
โ โโโ 100002.jpg
โ โโโ 100003.jpg
โโโ labels # xx.txt example
โโโ train2017
โ โโโ 000001.txt
โ โโโ 000002.txt
โ โโโ 000003.txt
โโโ val2017
โโโ 100001.txt
โโโ 100002.txt
โโโ 100003.txt
Auto LabelImg
Link ๏ผhttps://github.com/ppogg/AutoLabelImg
You can use LabelImg based YOLOv5-5.0 and YOLOv5-Lite to AutoAnnotate, biubiubiu ๐ ๐ ๐
Model Hub
Here, the original components of YOLOv5 and the reproduced components of YOLOv5-Lite are organized and stored in the model hub๏ผ
ncnn for arm-cpu
mnn for arm-cpu
openvino x86-cpu or x86-vpu
tensorrt(C++) for arm-gpu or arm-npu or x86-gpu
tensorrt(Python) for arm-gpu or arm-npu or x86-gpu
Android for arm-cpu
This is a Redmi phone, the processor is Snapdragon 730G, and yolov5-lite is used for detection. The performance is as follows:
Android_v5Lite-s: https://drive.google.com/file/d/1CtohY68N2B9XYuqFLiTp-Nd2kuFWgAUR/view?usp=sharing
Android_v5Lite-g: https://drive.google.com/file/d/1FnvkWxxP_aZwhi000xjIuhJ_OhqOUJcj/view?usp=sharing
new android app:[link] https://pan.baidu.com/s/1PRhW4fI1jq8VboPyishcIQ [keyword] pogg
What is YOLOv5-Lite S/E model: zhihu link (Chinese): https://zhuanlan.zhihu.com/p/400545131
What is YOLOv5-Lite C model: zhihu link (Chinese): https://zhuanlan.zhihu.com/p/420737659
What is YOLOv5-Lite G model: zhihu link (Chinese): https://zhuanlan.zhihu.com/p/410874403
How to deploy on ncnn with fp16 or int8: csdn link (Chinese): https://blog.csdn.net/weixin_45829462/article/details/119787840
How to deploy on mnn with fp16 or int8: zhihu link (Chinese): https://zhuanlan.zhihu.com/p/672633849
How to deploy on onnxruntime: zhihu link (Chinese): https://zhuanlan.zhihu.com/p/476533259(old version)
How to deploy on tensorrt: zhihu link (Chinese): https://zhuanlan.zhihu.com/p/478630138
How to optimize on tensorrt: zhihu link (Chinese): https://zhuanlan.zhihu.com/p/463074494
https://github.com/ultralytics/yolov5
https://github.com/megvii-model/ShuffleNet-Series
https://github.com/Tencent/ncnn
If you use YOLOv5-Lite in your research, please cite our work and give a star โญ:
@misc{yolov5lite2021,
title = {YOLOv5-Lite: Lighter, faster and easier to deploy},
author = {Xiangrong Chen and Ziman Gong},
doi = {10.5281/zenodo.5241425}
year={2021}
}