haotian/duan8
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
297 Commits 2 Branches 0 Tags 7.8 MiB
f2ac76df5a
Commit Graph

297 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Subtain Malik
f2ac76df5a
added MLP with proper documentation (#874) 
* [MLP]: create using core TensortRT python APIs

* [MLP]: README.md added with chart

* [MLP]: C++ TensorRT APIs added

* [MLP]: Updated Docs

* [MLP]: basic .wts added with APIs

* [MLP]: Updated minor details

* [MLP]: fix
 2022-01-10 11:34:23 +08:00
Vijaysrinivas Rajagopal
e6b22917e6
YOLOv3 TRT Python (#873) 
* change yolov5_trt to fit yolov3 output

* change ref from yolov5 to v3
 2022-01-07 11:09:48 +08:00
xupengao
5f1b048a29
fix bug (#808) 
* Add files via upload

* check_file

* remove invalid data

* remove data

* Optimization and modification

add all  yolo model

* fix bug

* fix bug

Co-authored-by: Wang Xinyu <wangxinyu_es@163.com>
 2021-11-09 13:53:20 +08:00
Silmeria112
befdf6ee64
update gen_wts.py for yolov5 to support custom model with different anchor from official model(#788) 2021-10-29 11:10:12 +08:00
Mr.Li
4697890416
Fix error when using yolov5-v6 custom mode (#782) 2021-10-26 13:25:30 +08:00
Jason
4686140d25
yolov5: fix bbox inaccurate issue (#779) 
Co-authored-by: chenjiansong <chenjiansong@trunk.tech>
 2021-10-21 22:24:36 +08:00
liuqi
2fef5f0ef2
yolov5: read input images only once (#776) 
you only read once。 把预处理图片缓存下来，画框时不必读一次，一张图片后处理时间少10ms。
 2021-10-21 11:14:24 +08:00
wang-xinyu
e785f51d10 update readme 2021-10-20 05:31:24 +00:00
wang-xinyu
542ab93fd3 reformat preprocess code of yolov5 2021-10-19 11:49:37 +00:00
liuqi
e7a5f0acbc
preprocess image in GPU for yolov5 (#770) 
* Update CMakeLists.txt

add nvcc compile option ;
add source file : preprocess.cu

* Create preprocess.h

* Update README.md

Add some configurations

* Update README.md

* Update CMakeLists.txt

* Update CMakeLists.txt

* Create preprocess.cu

* Update yolov5.cpp

Preprocess the image in GPU;
Use pinned memory to speed up the copy;
Only "cudamemcpy" once in function "doInference";

* Update preprocess.h

* Update CMakeLists.txt

* Update preprocess.cu

* Update yolov5.cpp

* Update preprocess.cu

* Update preprocess.cu

* Update preprocess.cu

* Update preprocess.cu

* Update preprocess.cu

* Update preprocess.cu

* Update preprocess.cu
 2021-10-19 18:56:47 +08:00
wang-xinyu
ae329092a1 update yolov5 readme 2021-10-18 07:43:48 +00:00
Wang Xinyu
8aa36a36f6
Update common.hpp 2021-10-18 13:51:44 +08:00
xupengao
a01fae5022
yolov6 6.0 Modification and optimization (#764) 
* Add files via upload

* check_file

* remove invalid data

* remove data

* Optimization and modification

add all  yolo model

Co-authored-by: Wang Xinyu <wangxinyu_es@163.com>
 2021-10-18 13:47:08 +08:00
xupengao
bcc01facfa
yolov5 6.0 (#763) 
* Add files via upload

* check_file

* remove invalid data

* remove data
 2021-10-18 11:20:23 +08:00
wang-xinyu
791c8a4484 update yolov5 readme 2021-10-13 12:15:10 +00:00
JumpPandaer
c70a338827
fix a bug with maskrcnn (#758) 2021-10-13 19:04:39 +08:00
liufqing
7c1a145c34
add resnet34 (#741) 
* hello

* add resnet34
 2021-09-30 15:18:04 +08:00
Armassarion
9fefbed77f
Fix a potential bug in detection number reset between each inference (#733) 
Co-authored-by: Ksander <ksander.ejjaaouani@digeiz.com>
 2021-09-22 19:22:28 +08:00
HaiyangP
f7d224f9e4
Add the coordinate clipping op before nms (#704) 
* add the coordinate clipping op before nms

* modified the pytorch related codes

* modified the pytorch related codes
 2021-09-02 19:35:30 +08:00
Jian Han Lim
0e254088a9
Update retinaface_trt.py, preprocess_image code (#703) 
Proposed solution to solve this issue #702
 2021-09-01 14:29:55 +08:00
Wang Xinyu
35a3301bf7
Update README.md 2021-08-31 19:10:16 +08:00
FamousDirector
9f9109fdc9
give retinaface TRT8 support (#698) 
* give retinaface TRT8 support

* add macros.h to retinaface conversion

* add macros.h to retinaface conversion
 2021-08-31 19:05:43 +08:00
HaiyangP
7c4476f23c
A pure numpy version for yolov5_trt.py (#700) 
* This code provides a pure numpy manner to write nms operation in yolov5_trt.py. I think it can offer simple and convenient TensorRT experience for yolov5

* delete torch related codes
 2021-08-31 11:32:34 +08:00
wang-xinyu
32049c61d1 update yolov5 readme 2021-08-30 05:52:02 +00:00
Cristi Fati
bbbe455a45
Improvements (generate .wts file) (#691) 
* Improve output (wts) file name generation

* Named arguments

* Minor: empty out argument
 2021-08-30 11:14:25 +08:00
HaiyangP
bd0d13f610
add the tensorrt python api for hrnet (#692) 2021-08-27 20:32:54 +08:00
JumpPandaer
ec20cead6f
guide for measuring performance (#676) 2021-08-17 10:50:43 +08:00
JumpPandaer
ccdf0e6316
update README (#673) 
add latency

improve nms
 2021-08-16 10:59:57 +08:00
wang-xinyu
a34527a48d create macros.h for yolov5 2021-08-05 11:57:48 +00:00
BaofengZan
55ad6259ce
Fix compilation problems on Windows (#644) 2021-07-30 10:57:29 +08:00
JumpPandaer
abcc753b19
add detr-res101 support (#645) 
* add detr

* Update README.md

* add int8 quantization

fix some known bugs

* add detr-res101 support

update detr README.md

update rcnn README.md
 2021-07-26 00:37:42 +08:00
Jow Xiong
f81d5ba7f4
improperly use about stl at line 35 (#621) 2021-07-22 10:51:16 +08:00
Huang ZhuangZhuang
2e53f0cae7
update infer preprocessing (#623) 
There are some problems with the preprocessing on LPRnet when infer, which will cause the predicted output results to be inconsistent with the pytorch version; as shown, the preprocessing has been updated here
 2021-07-15 12:46:18 +08:00
Aymeric Dujardin
b7a754e2de
Fix for TensorRT 8 support (#540) 2021-07-05 12:02:56 +08:00
JumpPandaer
9bb2f6771c
add int8 quantization (#612) 
* add detr

* Update README.md

* add int8 quantization

fix some known bugs
 2021-07-05 11:15:36 +08:00
wang-xinyu
14e220f421 fix yolov5 get_depth round bug 2021-07-01 11:57:06 +00:00
Wang Xinyu
917ad809ae
Update README.md 2021-07-01 13:14:36 +08:00
JumpPandaer
161e6b05ee
add detr (#606) 
* add detr

* Update README.md
 2021-07-01 13:09:51 +08:00
JumpPandaer
7b85cfd392
fix bugs (#602) 
see note in README
 2021-06-28 10:31:59 +08:00
wang-xinyu
101e60348e fix ofstream binary issue for windows #585 2021-06-16 07:58:13 +00:00
Wang Xinyu
f8c537586a
Update README.md 2021-06-11 10:42:16 +08:00
weiwei zhou
93e728769d
Add EfficientNet (#590) 
* create psenet

create psenet with weight from tensorflow

* delete some useless code

* repalce tab with 4 blanks

* fix network bug, rewrite post-processing pse algorithm

* update readme

* update readme

* add RepVGG

* fix typo

* add hrnetseg w18 w32 w48

* add hrnetseg with ocr w18 w32 w48

* merge hrnet and small, add hrnet_ocr

* fix warning

* change project name

* add efficientnet
 2021-06-11 10:34:40 +08:00
makaveli
668d89bbd5
fix: yolov5 get width (#566) 2021-05-25 20:47:16 +08:00
ybw108
d274d90722
Fix a bug of arcface-mobilefacenet (#562) 
Fix a bug which can cause different output results from tensorrt and insightface(mxnet) arcface-mobilefacenet. Change the `eps` value of `BatchNorm2d` layer in `conv_bn_relu()`, `conv_bn()` and `DepthWise()` from `2e-5` to `1e-3` according to the origin mxnet implementation of mobilefacenet from insightface.
 2021-05-24 18:31:41 +08:00
wang-xinyu
31646b6690 update readme 2021-05-24 03:49:13 +00:00
SsisyphusTao
d9e2cbc294
add centernet dla34 ctdet task. (#558) 
* add centernet dla34 ctdet task.

* update readme.

* update readme and fix a bug in sample.

Co-authored-by: chandler <chandler@invix.com>
 2021-05-23 16:29:14 +08:00
irvingzhang0512
1457e6fcb7
Fix a bug of arcface-mobilefacenet output size (#556) 2021-05-21 10:55:22 +08:00
wang-xinyu
4866607082 update readme 2021-05-17 11:44:48 +00:00
ybw108
53ba7e97b1
Add: arcface-mobilefacenet (#545) 
* add arcface-mobilefacenet

* Update arcface-mobilefacenet.cpp

* modify CMmakeLists.txt, add arcface-r100.cpp

* fix typos,resolve conflicts,update README.md

Co-authored-by: Admin <huangz11@chinatelecom.cn>
 2021-05-17 18:56:41 +08:00
Huang ZhuangZhuang
20fce2bce4
Update retina_r50.cpp and retina_nmt.cpp (#546) 2021-05-16 11:31:33 +08:00