文件列表:

Config.txt (228, 2016-11-17)
GoogLeNetFWD.asv (25860, 2016-11-17)
GoogLeNetFWD.m (27296, 2016-11-17)
Intermed_Results (0, 2016-11-17)
Intermed_Results\100_inception_5a_3x3_reduce.mat (13734, 2016-11-17)
Intermed_Results\101_inception_5a_3x3.mat (14574, 2016-11-17)
Intermed_Results\102_inception_5a_5x5_reduce.mat (3507, 2016-11-17)
Intermed_Results\103_inception_5a_5x5.mat (4198, 2016-11-17)
Intermed_Results\104_inception_5a_pool.mat (28155, 2016-11-17)
Intermed_Results\105_inception_5a_pool_proj.mat (4966, 2016-11-17)
Intermed_Results\106_inception_5a_output.mat (37150, 2016-11-17)
Intermed_Results\107_inception_5a_output_inception_5a_output_0_split_0.mat (37150, 2016-11-17)
Intermed_Results\108_inception_5a_output_inception_5a_output_0_split_1.mat (37150, 2016-11-17)
Intermed_Results\109_inception_5a_output_inception_5a_output_0_split_2.mat (37150, 2016-11-17)
Intermed_Results\10_pool2_3x3_s2_pool2_3x3_s2_0_split_1.mat (216553, 2016-11-17)
Intermed_Results\110_inception_5a_output_inception_5a_output_0_split_3.mat (37150, 2016-11-17)
Intermed_Results\111_inception_5b_1x1.mat (16875, 2016-11-17)
Intermed_Results\112_inception_5b_3x3_reduce.mat (8916, 2016-11-17)
Intermed_Results\113_inception_5b_3x3.mat (14769, 2016-11-17)
Intermed_Results\114_inception_5b_5x5_reduce.mat (3548, 2016-11-17)
Intermed_Results\115_inception_5b_5x5.mat (5864, 2016-11-17)
Intermed_Results\116_inception_5b_pool.mat (29463, 2016-11-17)
Intermed_Results\117_inception_5b_pool_proj.mat (3909, 2016-11-17)
Intermed_Results\118_inception_5b_output.mat (40794, 2016-11-17)
Intermed_Results\119_pool5_7x7_s1.mat (2945, 2016-11-17)
Intermed_Results\11_pool2_3x3_s2_pool2_3x3_s2_0_split_2.mat (216553, 2016-11-17)
Intermed_Results\120_loss3_classifier.mat (3931, 2016-11-17)
Intermed_Results\121_prob.mat (3976, 2016-11-17)
Intermed_Results\12_pool2_3x3_s2_pool2_3x3_s2_0_split_3.mat (216553, 2016-11-17)
Intermed_Results\13_inception_3a_1x1.mat (96173, 2016-11-17)
Intermed_Results\14_inception_3a_3x3_reduce.mat (152381, 2016-11-17)
Intermed_Results\15_inception_3a_3x3.mat (148800, 2016-11-17)
Intermed_Results\16_inception_3a_5x5_reduce.mat (29960, 2016-11-17)
Intermed_Results\17_inception_3a_5x5.mat (41966, 2016-11-17)
Intermed_Results\18_inception_3a_pool.mat (152830, 2016-11-17)
Intermed_Results\19_inception_3a_pool_proj.mat (39980, 2016-11-17)
Intermed_Results\1_data.mat (553391, 2016-11-17)
Intermed_Results\20_inception_3a_output.mat (326171, 2016-11-17)
Intermed_Results\21_inception_3a_output_inception_3a_output_0_split_0.mat (326171, 2016-11-17)
... ...

## Forward Implementation of GoogLeNet In this project, we implement the forward path of GoogLeNet in MATLAB. The required parameters and intermediate results are included in the project. The project doesn't need Caffe or any external library. ## How to run 1. Download the project 2. Open GoogLeNetFWD.m in MATLAB and run it The projects will be executed with a default input. You should see the accuracy numbers (difference between the current implementation and default implementation). ## Input If you want to change the input image, change the file address (input_file) in Confix.txt file. Please notice that when you change the input image, you should set the (cmp) parameter (in the Confix.txt) file to zero. (Why? The intermediate results which are available in the project are for the default image. If you change the default image, you cannot compare intermediate results. However, the network output is always correct.) ## Contact mmotamedi@ucdavis.edu ## Citing This project is devloped as a part of the following research. If the project helped your research, please kindly cite the paper: ``` Mohammad Motamedi, Philipp Gysel, Venkatesh Akella and Soheil Ghiasi, “Design Space Exploration of FPGA-Based Deep Convolutional Neural Network”, IEEE/ACM Asia-South Pacific Design Automation Conference (ASPDAC), January 2016. ```

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：