haotian/duan8
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
466 Commits 2 Branches 0 Tags 7.8 MiB
3f76507a9f
Commit Graph

7 Commits

Author SHA1 Message Date
HaiyangP
877bb24389
Optional nms methods supporting for rcnn (#1274) 
* Support to output the instance segmentation results on an original input image, rather than on a padded one.

* modifications according to freedenS's suggestions

* modification for padding

* merge preprocessImg and preprocessImg_ and delete all the debug codes.

* bug of reference passing

* make the vars consistent with the previous rcnn.cpp and update the README

* support different nms methods

* cxx11

* add contributors and delete interpretations

* update README

* fix some problems

* fix the bugs

---------

Co-authored-by: benkang wang <benkang.wang@uisee.com>
 2023-04-19 17:00:14 +08:00
nengwp
f2521a5ffa
RCNN upgrade to support TensorRT 7&8 (#1256) 
* rcnn upgrade to support TensorRT 8.

* Update RpnDecodePlugin.h

Remove Chinese

* Update rcnn.cpp

Remove Chinese

* Update backbone.hpp

Remove Chinese

* Update backbone.hpp

* Update rcnn.cpp

* Update rcnn.cpp

* Update rcnn.cpp

* Update MaskRcnnInferencePlugin.h

* rcnn upgrade to support TensorRT 8.x

* rcnn upgrade to support TensorRT 8.x

* Update macros.h

* Update README.md

---------

Co-authored-by: nengwp <nengwp@github.ai>
Co-authored-by: Wang Xinyu <shaywxy@gmail.com>
 2023-03-07 19:26:53 +08:00
HaiyangP
6636cc26fd
Modifications in rcnn.cpp for more applicable outputs of instance segmentation. (#946) 
* Support to output the instance segmentation results on an original input image, rather than on a padded one.

* modifications according to freedenS's suggestions

* modification for padding

* merge preprocessImg and preprocessImg_ and delete all the debug codes.

* bug of reference passing

* make the vars consistent with the previous rcnn.cpp and update the README

Co-authored-by: benkang wang <benkang.wang@uisee.com>
 2022-06-06 11:03:28 +08:00
JumpPandaer
2c4fdea67d
update rcnn (#521) 
* add MaskRcnn(C4)

* add MaskRcnnInference plugin for mask selecting

* split ROIHeads to BOXHead and MaskHead

* remove unuseful parameters in createEngine_rcnn and BuildRcnnModel

* change the type of scores_h, boxes_h and classes_h from unique_ptr to vector

* add doInference

* add maskrcnn postprocess

* update README.md

* update rcnn

* fix bugs with R18 and R34
 add BasicBlock for R18 and R34
 add STRIDE_IN_1X1, MakeStage is same with detectron2 now.

* update README.md

* update rcnn

replace picture with url
 2021-05-07 10:50:55 +08:00
JumpPandaer
8d8e2c9b0c
add MaskRcnn(C4) (#498) 
* add MaskRcnnInference plugin for mask selecting

* split ROIHeads to BOXHead and MaskHead

* remove unuseful parameters in createEngine_rcnn and BuildRcnnModel

* change the type of scores_h, boxes_h and classes_h from unique_ptr to vector

* add doInference

* add maskrcnn postprocess

* update README.md
 2021-04-23 10:42:22 +08:00
JumpPandaer
5d6b09cf5a
update rcnn (#495) 
add pixel_std in preprocess
fix synchronization error in roialign and cudaMemcpyAsync
improve coding style, limit line length less than 120 and so on
update README.md
upgrade TensorRT to 7.2
 2021-04-21 10:55:08 +08:00
JumpPandaer
891202d458
add FasterRcnn(R50C4) (#475) 
* Update yolov5.cpp

implement upsample with IResizeLayer

* Update yolov5.cpp

repair mistake and test it

* add FasterRcnn(R50C4)

maskrcnn is working

* add FasterRcnn(R50C4)

adjust indent

* update README

adjust coding style
 2021-04-16 11:41:18 +08:00