lewes6369 / Tensorrt Yolov3
Licence: mit
TensorRT for Yolov3
Stars: ✭ 451
Projects that are alternatives of or similar to Tensorrt Yolov3
Caffe Yolov3
A real-time object detection framework of Yolov3/v4 based on caffe
Stars: ✭ 435 (-3.55%)
Mutual labels: caffe, yolov3
Mobilenet Yolo
A caffe implementation of MobileNet-YOLO detection network
Stars: ✭ 825 (+82.93%)
Mutual labels: caffe, yolov3
Darknet2caffe
Convert darknet weights to caffemodel
Stars: ✭ 127 (-71.84%)
Mutual labels: caffe, yolov3
Caffe Yolov3 Windows
A windows caffe implementation of YOLO detection network
Stars: ✭ 210 (-53.44%)
Mutual labels: caffe, yolov3
Invoice
增值税发票OCR识别,使用flask微服务架构,识别type:增值税电子普通发票,增值税普通发票,增值税专用发票;识别字段为:发票代码、发票号码、开票日期、校验码、税后金额等
Stars: ✭ 381 (-15.52%)
Mutual labels: yolov3
Drbox
A deep learning based algorithm to detect rotated object, for example, objects in remote sensing images
Stars: ✭ 379 (-15.96%)
Mutual labels: caffe
Pytorch Caffe
load caffe prototxt and weights directly in pytorch
Stars: ✭ 407 (-9.76%)
Mutual labels: caffe
Normface
NormFace: L2 HyperSphere Embedding for Face Verification, 99.21% on LFW
Stars: ✭ 359 (-20.4%)
Mutual labels: caffe
Gocv
Go package for computer vision using OpenCV 4 and beyond.
Stars: ✭ 4,511 (+900.22%)
Mutual labels: caffe
Vpgnet
VPGNet: Vanishing Point Guided Network for Lane and Road Marking Detection and Recognition (ICCV 2017)
Stars: ✭ 382 (-15.3%)
Mutual labels: caffe
Seefood
Inspired by HBO's Silicon Valley: SeeFood is an iOS app that uses CoreML to detect various dishes
Stars: ✭ 445 (-1.33%)
Mutual labels: caffe
Mini Caffe
Minimal runtime core of Caffe, Forward only, GPU support and Memory efficiency.
Stars: ✭ 373 (-17.29%)
Mutual labels: caffe
Visualdl
Deep Learning Visualization Toolkit(『飞桨』深度学习可视化工具 )
Stars: ✭ 4,258 (+844.12%)
Mutual labels: caffe
Mobilenetv2 Ssdlite
Caffe implementation of SSD and SSDLite detection on MobileNetv2, converted from tensorflow.
Stars: ✭ 435 (-3.55%)
Mutual labels: caffe
Centerx
This repo is implemented based on detectron2 and centernet
Stars: ✭ 403 (-10.64%)
Mutual labels: caffe
Trainyourownyolo
Train a state-of-the-art yolov3 object detector from scratch!
Stars: ✭ 399 (-11.53%)
Mutual labels: yolov3
TRTForYolov3
Desc
tensorRT for Yolov3
Test Enviroments
Ubuntu 16.04
TensorRT 5.0.2.6/4.0.1.6
CUDA 9.2
Models
Download the caffe model converted by official model:
If run model trained by yourself, comment the "upsample_param" blocks, and modify the prototxt the last layer as:
layer {
#the bottoms are the yolo input layers
bottom: "layer82-conv"
bottom: "layer94-conv"
bottom: "layer106-conv"
top: "yolo-det"
name: "yolo-det"
type: "Yolo"
}
It also needs to change the yolo configs in "YoloConfigs.h" if different kernels.
Run Sample
#build source code
git submodule update --init --recursive
mkdir build
cd build && cmake .. && make && make install
cd ..
#for yolov3-608
./install/runYolov3 --caffemodel=./yolov3_608.caffemodel --prototxt=./yolov3_608.prototxt --input=./test.jpg --W=608 --H=608 --class=80
#for fp16
./install/runYolov3 --caffemodel=./yolov3_608.caffemodel --prototxt=./yolov3_608.prototxt --input=./test.jpg --W=608 --H=608 --class=80 --mode=fp16
#for int8 with calibration datasets
./install/runYolov3 --caffemodel=./yolov3_608.caffemodel --prototxt=./yolov3_608.prototxt --input=./test.jpg --W=608 --H=608 --class=80 --mode=int8 --calib=./calib_sample.txt
#for yolov3-416 (need to modify include/YoloConfigs for YoloKernel)
./install/runYolov3 --caffemodel=./yolov3_416.caffemodel --prototxt=./yolov3_416.prototxt --input=./test.jpg --W=416 --H=416 --class=80
Performance
Model | GPU | Mode | Inference Time |
---|---|---|---|
Yolov3-416 | GTX 1060 | Caffe | 54.593ms |
Yolov3-416 | GTX 1060 | float32 | 23.817ms |
Yolov3-416 | GTX 1060 | int8 | 11.921ms |
Yolov3-608 | GTX 1060 | Caffe | 88.489ms |
Yolov3-608 | GTX 1060 | float32 | 43.965ms |
Yolov3-608 | GTX 1060 | int8 | 21.638ms |
Yolov3-608 | GTX 1080 Ti | float32 | 19.353ms |
Yolov3-608 | GTX 1080 Ti | int8 | 9.727ms |
Yolov3-416 | GTX 1080 Ti | float32 | 9.677ms |
Yolov3-416 | GTX 1080 Ti | int8 | 6.129ms |
Eval Result
run above models with appending --evallist=labels.txt
int8 calibration data made from 200 pics selected in val2014 (see scripts dir)
Model | GPU | Mode | dataset | MAP(0.50) | MAP(0.75) |
---|---|---|---|---|---|
Yolov3-416 | GTX 1060 | Caffe(fp32) | COCO val2014 | 50.33 | 33.00 |
Yolov3-416 | GTX 1060 | float32 | COCO val2014 | 50.27 | 32.98 |
Yolov3-416 | GTX 1060 | int8 | COCO val2014 | 44.15 | 30.24 |
Yolov3-608 | GTX 1060 | Caffe(fp32) | COCO val2014 | 52.89 | 35.31 |
Yolov3-608 | GTX 1060 | float32 | COCO val2014 | 52.84 | 35.26 |
Yolov3-608 | GTX 1060 | int8 | COCO val2014 | 48.55 | 35.53 |
Notice:
- caffe implementation is little different in yolo layer and nms, and it should be the similar result compared to tensorRT fp32.
Details About Wrapper
see link TensorRTWrapper
Note that the project description data, including the texts, logos, images, and/or trademarks,
for each open source project belongs to its rightful owner.
If you wish to add or remove any projects, please contact us at [email protected].