nihui/ncnn - ncnn - 开源协同云脑生态支撑系统

Commit Graph

Author	SHA1	Message	Date
nihui	3ef995ed1e	format code style and setup restyled.io (#1840 )	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
nihui	9929d52885	less duplicated code for crop layer, slice axes starts from 0	6 years ago
kalcohol	01853010ea	type conversion and MSVC _CRT_SECURE_NO_DEPRECATE warning fix (#1524 ) * fix type conversion warning * fix MSVC _CRT_SECURE_NO_DEPRECATE warning * type conversion: missing one * fix MSVC _CRT_SECURE_NO_DEPRECATE warning and type warning * reformat * convert google::protobuf::int64 to size_t to avoid type impl between msvc and gcc gap * remove C++11 auto * tab -> space	6 years ago
nihuini	c9a9128db7	convert caffe Slice slice_dim	6 years ago
nihuini	7d3ebe2d69	%g may output number without dot, stick to %e	6 years ago
nihuini	d67d2531ae	convert some well-known noop, printf %g to preserve precision	6 years ago
tpoisonooo	8dbafe7764	constraint input value to [-127, +127] (#1258 ) * constraint input value to [-127, +127] * keep new line at the end	6 years ago
nihui	7ab968e6e1	fix gpu crop, convert crop offset with axis	7 years ago
nihuini	9787625e4b	warn users about the old wrong softmax behavior on axis not zero	7 years ago
BUG1989	df3d224484	new int8 implement,better accuracy (#749 ) * add the armv7a conv3x3s1 implement without overflow,remove old codes * fix the bug of conv3x3s2 packed int8 * new int8 implement,weight quant by perchanel,better accuracy~ * fix the bug of conv3x3s1 packed int8 neon * add the naive c fp32 and int8 winograd F(2,3) * add the neon intrinsic int8 winograd F(2,3) * optimize the armv7a int8 winograd F(2,3) with neon assembly * optimize the armv7a int8 winograd F(2,3) input transform with assembly. * add the requantize layer and int8 relu implement. * add graph optimize conv1x1s2 -> conv1x1s1,begin optimize int8 aarch64. * fix int8 bugs * add the c naive im2col with sgemm * add aarch64 int8 winograd f23, conv3x3s2 naive implement * add the int8 sgemm conv7x7s2 on x86/armv7a platform * optimize the int8 sgemm by neon intrinsic and packed kernel * optimize the int8 sgemm with packed data * optimize the int8 sgemm with armv7a neon assembly * add the int8 sgemm on arm64-v8a platform * perpare to merge latest codes from master * add the int8 param files * In the Class Net,add the fuse_network method	7 years ago
nihui	d85775fbcd	fix softmax axis order on 3-dim, fix caffe reshape conversion, regenerate ssd param	7 years ago
nihuini	9b10729860	fix read_int8scale_table, second try	7 years ago
nihuini	501cdf2adf	fscanf may advance stream position, workaround it	7 years ago
nihuini	83818bd52e	convert roialign	7 years ago
nihuini	faac207410	convert psroipooling	7 years ago
Ming-Yang Liu	dfbed5a0ad	Support yolov3 and mobilenet yolov3	7 years ago
BUG1989	d5df2301cb	fix the caffe2ncnn that "DepthwiseConvolution" loss group num param (#540 ) * fix the caffe2ncnn bug that "DepthwiseConvolution" loss group num param	7 years ago
nihuini	2dbaf6f7b7	store int8 scale in binary	7 years ago
nihui	5d04a3a45c	layer holds bottom blob scale, depthwise convolution read group scales	7 years ago
nihui	354b95256c	bump param version, backward compatible	7 years ago
nihui	a169cec363	core int8 inference, quantize and dequantize, net using flag, caffe2ncnn reads int8 scale table	7 years ago
nihui	0b6791e2ba	convert BN ReLU6 Reorg YoloDetectionOutput Embed LSTM	8 years ago
nihuini	588487a8a0	convert caffe crop layer with three offset, fix #165	8 years ago
nihuini	babbb604e1	fix deconvolution weight order	8 years ago
nihuini	3b8a3f6764	convert depthwiseconvolution from yonghenglh6 branch	8 years ago
nihuini	b560252af1	handle batchnorm scale_factor zero, fix #302	8 years ago
nihuini	db4b04da53	fix slice with implicit slice point	8 years ago
nihui	0e41c37250	convert convolution pad_w pad_h	8 years ago
nihui	d2c01019aa	fix convert depthwise deconvolution, fix #300	8 years ago
yqz	c6abc872df	fix mvn no load param in caffe2ncnn	8 years ago
nihui	c9789fb879	slice dim	8 years ago
nihui	0f52418023	change input param order to w h c, replace caffe MemoryData to Input	8 years ago
nihui	abfd3ea6c8	convert non-square convolution pooling param	8 years ago
Zexin, Hu	81fb3818a5	add ShuffleChannel layer, only cpp, no arm yet (#210 ) * add ShuffleChannel layer, only cpp, no arm yet * add assign	8 years ago
nihuini	25f19c2009	implement external scale blob, support SENet	8 years ago
nihuini	96aa5d61af	do not print relu negative slope if it is zero	8 years ago
nihuini	8e12bc916d	compability improvement for https://github.com/farmingyard/caffe-mobilenet	8 years ago
nihuini	e8248ecab0	check dropout has_scale_train	8 years ago
tedder59	4d59d0afda	Add depthwise Deconvolution. (#187 ) * add depthwise deconvolution. * add depthwise deconvolution. * fix some syntax error and uncessary modification	8 years ago
nihuini	20f49b677c	move caffe2ncnn to subdir, build mxnet2ncnn	8 years ago

41 Commits (bedf00a5edf85ae8af33bc72ebd85fe325da1271)