ruru5697
e58a35aa65
fix darknet2ncnn maxpool padding size ( #2038 )
* fix maxpool padding size
* restyle
5 years ago
Zhuo Zhang
0fc5ee1383
fix #2056 ( #2057 )
5 years ago
nihuini
63247a32a7
mlir fix keras conv2d dense bias, fix pooling type
5 years ago
nihuini
6a15cb1a0e
mlir fuse keras conv2d dense
5 years ago
nihuini
d279e1f514
mlir fuse instancenorm without affine
5 years ago
nihuini
78580dc648
sync tf dialect, mlir fuse instancenorm
5 years ago
nihuini
ea17a46b34
fuse memorydata split binaryop
5 years ago
nihui
1cb55454eb
fix onnx2ncnn resize crash with optional sizes
5 years ago
nihuini
7b3ab5024d
convert to gemm layer
5 years ago
nihuini
4b470949c2
update tf2 dialect, adapt new mlir api, convert tf.Maximum tf.Minimum tf.ResizeBilinear
5 years ago
nihuini
01d9ebf115
sometimes mxnet use either scalar or array in one attribute value, #1976
5 years ago
zchrissirhcz
7d2c3247bc
fix #1941 darknet2ncnn maxpool out_w issue ( #1972 )
* fix #1941 darknet2ncnn maxpool out_w issue
* fix typo caused build fail
* fix opencv4 c++11 flags in example folder
5 years ago
zchrissirhcz
b80b84fda5
fix #1542 ; fix avx2 uint16_t including ( #1968 )
* fix #1542 ; fix avx2 uint16_t including
for #1542 , it is for compatibility for opencv 2.x, such as on ubuntu 16.04 apt installed opencv
5 years ago
Tijmen Verhulsdonck
a3f7d2d520
Fix padding in ncnnoptimize not outputting all the params ( #1944 )
5 years ago
nihui
02e880fc42
fix concatv2 axis
5 years ago
nihuini
80bd7587e6
fuse normalize for simplified onnx
5 years ago
nihuini
d92e71722d
improve compability with onnx resize opset 10
5 years ago
nihuini
00ef566609
implement full permute tag support for reshape
5 years ago
nihuini
b2b86fea16
convert tf.ResizeNearestNeighbor
5 years ago
nihui
afb09cccbb
build mlir2ncnn with latest mlir with updated tf2 dialect ( #1925 )
5 years ago
nihui
46c6f3a4a2
hmmm, mlir need c++14 anyway
5 years ago
nihui
cfcee2298d
mlir2ncnn requires c+11
5 years ago
zhiliu6
cdbff653b8
Add yolov4 example option. ( #1913 )
* Add yolov4 example option.
Add yolov4-tiny for benchmark.
* Restyled by astyle
* Restyled by clang-format
* Restyled by astyle
Co-authored-by: Restyled.io <commits@restyled.io>
5 years ago
nihui
6e340f5e79
MLIR converter ( #1810 )
* tf2sb
* wip
* wip
* concat maxpool avgpool
* StridedSlice
* handle stridedslice mask
* convert mlir deconvolution
* convert deconvolution output shape
* convert tanh
* code format
5 years ago
zhiliu6
0bc45eed52
Support yolov4-tiny conversion. ( #1893 )
* Support yolov4-tiny conversion.
Add darknet2ncnn readme.
Add yolov4 example.
* Restyled by clang-format
* Restyled by astyle
* Restyled by clang-format
* Restyled by astyle
Co-authored-by: Restyled.io <commits@restyled.io>
5 years ago
nihuini
bcf982cc9d
fuse onnx gather after shufflechannel into split
6 years ago
nihuini
294b5b6e99
onnx fuse reverse style shufflechannel
6 years ago
nihuini
9238682778
onnx slice axes steps can be optional, #1850
6 years ago
nihui
3ef995ed1e
format code style and setup restyled.io ( #1840 )
6 years ago
mengfu188
f09ac067a5
update quantize doc ( #1831 )
6 years ago
JackieWu
ce2251db05
Improve ROIAlign (accelerate ROIAlign, support sampling ratio and aligned ROIAlign) ( #1820 )
* update roialign
* update tool for roialign
* fix ceil for roialign
* fix ios build
* fix
* fix
* make it compatibile with the old version
* fix ios build
* trigger CI
* fix test
* order
* build
* remove code
* merge roialign
* accelerate ROIAlign
* update note
* rename func
* roialign version
* trigger CI
* fix roialign
* use ref for pre-calc in roialign
* retrigger CI
* pre-alloc
6 years ago
zhiliu6
63d7e2c88d
Add support for darknet EfficientNetB0-Yolov3 conversion. ( #1821 )
* fuse mish and depthwise convolution
* darknet2ncnn: support enet-coco conversion.
6 years ago
nihuini
adf7736d8f
fuse onnx swish
6 years ago
zhiliu6
3bfabf1d6a
Add fused convolution and mish layer support. ( #1761 )
6 years ago
Guoxia Wang
7add0ea94e
fix ncnn2mem.cpp param bin save path ( #1766 )
6 years ago
nihuini
092d529d66
fuse rsub rdiv with scalar
6 years ago
nihuini
1c4d5bebc8
onnx Sub is BinaryOp
6 years ago
Naiyang Lin
ceef2470a5
Add logger.h ( #1753 )
6 years ago
nihuini
e6a04d1535
fuse_memorydata_binaryop fuse_binaryop_eltwise
6 years ago
zhiliu6
2a97606239
darknet2ncnn: Add merge output option. ( #1743 )
* darknet2ncnn: Add an option, which is enabled by default, to merge all yolo output into one layer. The convert result is compatible with yolov3 example and can be used in benchmark.
6 years ago
zhiliu6
585480220f
Add darknet to ncnn converter with YOLOv4 support. ( #1740 )
* Add darknet to ncnn converter with YOLOv4 support.
* darknet2ncnn: add copyright info, fix clang compatibility
* fix wrong target
* darknet2ncnn: compatible with clang++ 9.0.0
6 years ago
ZhangCY
8a3e9189df
mxnet2ncnn support list args in attrs ( #1741 )
Co-authored-by: zcy6735 <zcy6735@arcsoft.com.cn>
6 years ago
daquexian
949f6a4880
Fix bug of get_node_attr_from_input_ai ( #1699 )
6 years ago
nihuini
22f4212128
improve onnx hardsigmoid hardsiwsh fusion for fbnet
6 years ago
mengfu188
d06606b0cd
Tiny fix mxnet2ncnn tool ( #1691 )
6 years ago
nihui
88872f8ac3
fix int64 overflow, fix #1670
6 years ago
nihuini
3cbdd077e9
sometimes binaryop weight is missing in graph nodes, fix #1640
6 years ago
nihui
f3b39a3b4f
improve onnx hardsigmoid hardswish fusion, fix #1650 , ignore mips ci failure
6 years ago
nihuini
253e505765
improve compatibility with onnx slice opset version 10+
6 years ago
nihuini
9b09cc16b5
fix compability with onnx clip opset11, fix #1609
6 years ago