TDA2: ssd model import problem

khethan

Part Number: TDA2

Hi All,

I have already confirmed that it works well on TDA2p EVM as below when running tidl usecase using the already converted model bin(NET_OD/PRM_OD)

-. bbox : ok, display fps : ok(16fps)

However, I confirmed that the bin(NET_OD, PRM_OD) changed through tild_model_import.out does not work as it did before.

-. bbox: no, display fps : fail(8.4fps)

Below is the problem log and import file file. Can you help me with what is wrong?

modelImportIssue.zip

My development environment is as follows.

VSDK : 3.3.0.0

TIDLSRC: 1.1.1.0

bootmode ; SDBOOT

EVM: tda2p

protobuf ver : 3.2.0rc2

model: JDetNet

Thank you in advance.

BR,

Khethan

over 7 years ago

0 Praveen Eppa1 over 7 years ago

TI__Genius 17580 points

Hi Khethan,

Can you please check with new vision SDK release version 03.05.00.00, if you still face the issue please share the generated NET.bin and PARAM.bin from import tool for verification.

Thanks,
Praveen

0 khethan over 7 years ago in reply to Praveen Eppa1

Intellectual 470 points

Hi all

I am sorry, It's take a long time to write back.

I downloaded the version(3.5.0.0) you mentioned, but which is does not have a TIDL.

The stats_tool_out.bin file created using my import tool has been confirmed to work normally as below.

please refer my environment file.

I would appreciate any help on what is wrong.

modelImportIssue_2.zip

BR,

Khethan

0 Praveen Eppa1 over 7 years ago in reply to khethan

TI__Genius 17580 points

Hi Khethan,

I did not see any issue with your import tool files, because I just now imported and verified that it is detecting objects properly. I used the import executable (tidl_model_import.out.exe) from latest TIDL 01.01.02.00 release.

Can you please check with latest TIDL release?

Thanks,
Praveen

0 khethan over 7 years ago in reply to Praveen Eppa1

Intellectual 470 points

Hi Praveen,

I've just checked with the new version(1.1.2.0) you mentioned, but I still have problems.

please refer my log file.

t-1a84591e31034fac832d29ed8584666c1664508081_code-editor-fs">Fullscreen //e2e.ti.com/cfs-filesystemfile/__key/communityserver-discussions-components-files/791/import_5F00_tidl_5F00_1_5F00_1_5F00_2_5F00_0_5F00_log.txt?_=636770141620000000" import_tidl_1_1_2_0_log.txt">import_tidl_1_1_2_0_log.txt //e2e.ti.com/cfs-filesystemfile/__key/communityserver-discussions-components-files/791/import_5F00_tidl_5F00_1_5F00_1_5F00_2_5F00_0_5F00_log.txt?_=636770141620000000" import_tidl_1_1_2_0_log.txt">Download 84591e31034fac832d29ed8584666c1664508081_code-editor" 22px;min-width:1230px;margin:0px;" Windows [Version 6.1.7601] 2009 Microsoft Corporation. All rights reserved. _03_03_00_00\ti_components\algorithms\REL.TIDLSRC.01.01.02.00\modules\ti_dl\utils\tidlModel c:\PROCESSOR_SDK_VISION_03_03_00_00\ti_components\algorithms\REL.TIDLSRC.01.01.02 testvecs\config\import\tidl_import_JDetNet.txt File : d:\work\adas\TI\work_space\tidl\model\caffe_jacinto_models\trained\object_detection\voc0712\J P_dsFac_32_hdDS8_0\sparse\deploy.prototxt : d:\work\adas\TI\work_space\tidl\model\caffe_jacinto_models\trained\object_detection\voc0712\J P_dsFac_32_hdDS8_0\sparse\voc0712_ssdJacintoNetV2_iter_120000.caffemodel : ..\..\test\testvecs\config\tidl_models\jdetnet\tidl_net_jdetNet_ssd.bin : ..\..\test\testvecs\config\tidl_models\jdetnet\tidl_param_jdetNet_ssd.bin : ssdJacintoNetV2_deploy , data 0, -1 , 1 , x , x , x , x , 0 , 0 , 0 , 0 , 1 , 3 , 320 , 768 , 0 , data/bias 1, 1 , 1 , 0 , x , x , x , 1 , 3 , 320 , 768 , 1 , 3 , 320 , 768 , 737280 , conv1a 1, 1 , 1 , 1 , x , x , x , 1 , 3 , 320 , 768 , 1 , 32 , 160 , 384 , 147456000 , conv1b 1, 1 , 1 , 2 , x , x , x , 1 , 32 , 160 , 384 , 1 , 32 , 80 , 192 , 141557760 , res2a_branch2a 1, 1 , 1 , 3 , x , x , x , 1 , 32 , 80 , 192 , 1 , 64 , 80 , 192 , 283115520 , res2a_branch2b 1, 1 , 1 , 4 , x , x , x , 1 , 64 , 80 , 192 , 1 , 64 , 40 , 96 , 141557760 , res3a_branch2a 1, 1 , 1 , 5 , x , x , x , 1 , 64 , 40 , 96 , 1 , 128 , 40 , 96 , 283115520 , res3a_branch2b 1, 1 , 1 , 6 , x , x , x , 1 , 128 , 40 , 96 , 1 , 128 , 20 , 48 , 141557760 , res4a_branch2a 1, 1 , 1 , 7 , x , x , x , 1 , 128 , 20 , 48 , 1 , 256 , 20 , 48 , 283115520 , res4a_branch2b 1, 1 , 1 , 8 , x , x , x , 1 , 256 , 20 , 48 , 1 , 256 , 20 , 48 , 141557760 , pool4 1, 1 , 1 , 9 , x , x , x , 1 , 256 , 20 , 48 , 1 , 256 , 10 , 24 , 245760 , res5a_branch2a 1, 1 , 1 , 10 , x , x , x , 1 , 256 , 10 , 24 , 1 , 512 , 10 , 24 , 283115520 , res5a_branch2b 1, 1 , 1 , 11 , x , x , x , 1 , 512 , 10 , 24 , 1 , 512 , 10 , 24 , 141557760 , pool6 1, 1 , 1 , 12 , x , x , x , 1 , 512 , 10 , 24 , 1 , 512 , 5 , 12 , 122880 , pool7 1, 1 , 1 , 13 , x , x , x , 1 , 512 , 5 , 12 , 1 , 512 , 3 , 6 , 36864 , pool8 1, 1 , 1 , 14 , x , x , x , 1 , 512 , 3 , 6 , 1 , 512 , 2 , 3 , 12288 , ctx_output1 1, 1 , 1 , 9 , x , x , x , 1 , 256 , 20 , 48 , 1 , 256 , 20 , 48 , 62914560 , ctx_output2 1, 1 , 1 , 12 , x , x , x , 1 , 512 , 10 , 24 , 1 , 256 , 10 , 24 , 31457280 , ctx_output3 1, 1 , 1 , 13 , x , x , x , 1 , 512 , 5 , 12 , 1 , 256 , 5 , 12 , 7864320 , ctx_output4 1, 1 , 1 , 14 , x , x , x , 1 , 512 , 3 , 6 , 1 , 256 , 3 , 6 , 2359296 , ctx_output5 1, 1 , 1 , 15 , x , x , x , 1 , 512 , 2 , 3 , 1 , 256 , 2 , 3 , 786432 , ctx_output1/relu_mbox_loc 1, 1 , 1 , 16 , x , x , x , 1 , 256 , 20 , 48 , 1 , 16 , 20 , 48 , 3932160 , ctx_output1/relu_mbox_loc_perm 1, 1 , 1 , 21 , x , x , x , 1 , 16 , 20 , 48 , 1 , 1 , 1 , 15360 , 1 , ctx_output1/relu_mbox_conf 1, 1 , 1 , 16 , x , x , x , 1 , 256 , 20 , 48 , 1 , 84 , 20 , 48 , 20643840 , ctx_output1/relu_mbox_conf_perm 1, 1 , 1 , 23 , x , x , x , 1 , 84 , 20 , 48 , 1 , 1 , 1 , 80640 , 1 , ctx_output2/relu_mbox_loc 1, 1 , 1 , 17 , x , x , x , 1 , 256 , 10 , 24 , 1 , 24 , 10 , 24 , 1474560 , ctx_output2/relu_mbox_loc_perm 1, 1 , 1 , 26 , x , x , x , 1 , 24 , 10 , 24 , 1 , 1 , 1 , 5760 , 1 , ctx_output2/relu_mbox_conf 1, 1 , 1 , 17 , x , x , x , 1 , 256 , 10 , 24 , 1 , 126 , 10 , 24 , 7741440 , ctx_output2/relu_mbox_conf_perm 1, 1 , 1 , 28 , x , x , x , 1 , 126 , 10 , 24 , 1 , 1 , 1 , 30240 , 1 , ctx_output3/relu_mbox_loc 1, 1 , 1 , 18 , x , x , x , 1 , 256 , 5 , 12 , 1 , 24 , 5 , 12 , 368640 , ctx_output3/relu_mbox_loc_perm 1, 1 , 1 , 31 , x , x , x , 1 , 24 , 5 , 12 , 1 , 1 , 1 , 1440 , 1 , ctx_output3/relu_mbox_conf 1, 1 , 1 , 18 , x , x , x , 1 , 256 , 5 , 12 , 1 , 126 , 5 , 12 , 1935360 , ctx_output3/relu_mbox_conf_perm 1, 1 , 1 , 33 , x , x , x , 1 , 126 , 5 , 12 , 1 , 1 , 1 , 7560 , 1 , ctx_output4/relu_mbox_loc 1, 1 , 1 , 19 , x , x , x , 1 , 256 , 3 , 6 , 1 , 24 , 3 , 6 , 110592 , ctx_output4/relu_mbox_loc_perm 1, 1 , 1 , 36 , x , x , x , 1 , 24 , 3 , 6 , 1 , 1 , 1 , 432 , 1 , ctx_output4/relu_mbox_conf 1, 1 , 1 , 19 , x , x , x , 1 , 256 , 3 , 6 , 1 , 126 , 3 , 6 , 580608 , ctx_output4/relu_mbox_conf_perm 1, 1 , 1 , 38 , x , x , x , 1 , 126 , 3 , 6 , 1 , 1 , 1 , 2268 , 1 , ctx_output5/relu_mbox_loc 1, 1 , 1 , 20 , x , x , x , 1 , 256 , 2 , 3 , 1 , 16 , 2 , 3 , 24576 , ctx_output5/relu_mbox_loc_perm 1, 1 , 1 , 41 , x , x , x , 1 , 16 , 2 , 3 , 1 , 1 , 1 , 96 , 1 , ctx_output5/relu_mbox_conf 1, 1 , 1 , 20 , x , x , x , 1 , 256 , 2 , 3 , 1 , 84 , 2 , 3 , 129024 , ctx_output5/relu_mbox_conf_perm 1, 1 , 1 , 43 , x , x , x , 1 , 84 , 2 , 3 , 1 , 1 , 1 , 504 , 1 , mbox_loc 1, 5 , 1 , 22 , 27 , 32 , 37 , 1 , 1 , 1 , 15360 , 1 , 1 , 1 , 23088 , 1 , mbox_conf 1, 5 , 1 , 24 , 29 , 34 , 39 , 1 , 1 , 1 , 80640 , 1 , 1 , 1 , 121212 , 1 , detection_out 2, 2 , 1 , 46 , 47 , x , x , 1 , 1 , 1 , 23088 , 1 , 1 , 1 , 560 , 1 file .\tempDir\qunat_stats_config.txt ! , 0, -1 , 1 , x , x , x , x , x , x , x , x , 0 , 0 , 0 , , 1, 1 , 1 , 0 , x , x , x , x , x , x , x , 1 , 1 , 3 , 32 , 1, 1 , 1 , 1 , x , x , x , x , x , x , x , 2 , 1 , 3 , 32 , 1, 1 , 1 , 2 , x , x , x , x , x , x , x , 3 , 1 , 32 , 16 , 1, 1 , 1 , 3 , x , x , x , x , x , x , x , 4 , 1 , 32 , 8 , 1, 1 , 1 , 4 , x , x , x , x , x , x , x , 5 , 1 , 64 , 8 , 1, 1 , 1 , 5 , x , x , x , x , x , x , x , 6 , 1 , 64 , 4 , 1, 1 , 1 , 6 , x , x , x , x , x , x , x , 7 , 1 , 128 , 4 , 1, 1 , 1 , 7 , x , x , x , x , x , x , x , 8 , 1 , 128 , 2 , 1, 1 , 1 , 8 , x , x , x , x , x , x , x , 9 , 1 , 256 , 2 , 1, 1 , 1 , 9 , x , x , x , x , x , x , x , 10 , 1 , 256 , 2 , 1, 1 , 1 , 10 , x , x , x , x , x , x , x , 11 , 1 , 256 , 1 , 1, 1 , 1 , 11 , x , x , x , x , x , x , x , 12 , 1 , 512 , 1 , 1, 1 , 1 , 12 , x , x , x , x , x , x , x , 13 , 1 , 512 , 1 , 1, 1 , 1 , 13 , x , x , x , x , x , x , x , 14 , 1 , 512 , , 1, 1 , 1 , 14 , x , x , x , x , x , x , x , 15 , 1 , 512 , , 1, 1 , 1 , 9 , x , x , x , x , x , x , x , 16 , 1 , 256 , 2 , 1, 1 , 1 , 12 , x , x , x , x , x , x , x , 17 , 1 , 512 , 1 , 1, 1 , 1 , 13 , x , x , x , x , x , x , x , 18 , 1 , 512 , , 1, 1 , 1 , 14 , x , x , x , x , x , x , x , 19 , 1 , 512 , , 1, 1 , 1 , 15 , x , x , x , x , x , x , x , 20 , 1 , 512 , , 1, 1 , 1 , 16 , x , x , x , x , x , x , x , 21 , 1 , 256 , 2 , 1, 1 , 1 , 21 , x , x , x , x , x , x , x , 22 , 1 , 16 , 2 , 1, 1 , 1 , 16 , x , x , x , x , x , x , x , 23 , 1 , 256 , 2 , 1, 1 , 1 , 23 , x , x , x , x , x , x , x , 24 , 1 , 84 , 2 , 1, 1 , 1 , 17 , x , x , x , x , x , x , x , 26 , 1 , 256 , 1 , 1, 1 , 1 , 26 , x , x , x , x , x , x , x , 27 , 1 , 24 , 1 , 1, 1 , 1 , 17 , x , x , x , x , x , x , x , 28 , 1 , 256 , 1 , 1, 1 , 1 , 28 , x , x , x , x , x , x , x , 29 , 1 , 126 , 1 , 1, 1 , 1 , 18 , x , x , x , x , x , x , x , 31 , 1 , 256 , , 1, 1 , 1 , 31 , x , x , x , x , x , x , x , 32 , 1 , 24 , , 1, 1 , 1 , 18 , x , x , x , x , x , x , x , 33 , 1 , 256 , , 1, 1 , 1 , 33 , x , x , x , x , x , x , x , 34 , 1 , 126 , , 1, 1 , 1 , 19 , x , x , x , x , x , x , x , 36 , 1 , 256 , , 1, 1 , 1 , 36 , x , x , x , x , x , x , x , 37 , 1 , 24 , , 1, 1 , 1 , 19 , x , x , x , x , x , x , x , 38 , 1 , 256 , , 1, 1 , 1 , 38 , x , x , x , x , x , x , x , 39 , 1 , 126 , , 1, 1 , 1 , 20 , x , x , x , x , x , x , x , 41 , 1 , 256 , , 1, 1 , 1 , 41 , x , x , x , x , x , x , x , 42 , 1 , 16 , , 1, 1 , 1 , 20 , x , x , x , x , x , x , x , 43 , 1 , 256 , , 1, 1 , 1 , 43 , x , x , x , x , x , x , x , 44 , 1 , 84 , , 1, 5 , 1 , 22 , 27 , 32 , 37 , 42 , x , x , x , 46 , 1 , 1 , , 1, 5 , 1 , 24 , 29 , 34 , 39 , 44 , x , x , x , 47 , 1 , 1 , , 1, 2 , 1 , 46 , 47 , x , x , x , x , x , x , 49 , 1 , 1 , , 0, 1 , -1 , 49 , x , x , x , x , x , x , x , 0 , 1 , 1 , ,inBlkHeight ,inBlkPitch ,outBlkWidth ,outBlkHeight,outBlkPitch ,numInChs ,numOutChs ,numLclOutChs,numProcItrs ,numAccItrs ,numHorBlock ,numVerBlock ,inBlkChPitch,outBlkC 72 72 32 32 32 3 32 8 1 3 12 5 5184 1024 34 40 32 32 32 8 8 8 1 2 12 5 1360 1024 22 40 32 20 32 32 64 8 1 4 6 4 880 640 22 40 32 20 32 16 16 8 1 2 6 4 880 640 22 40 32 20 32 64 128 8 1 8 3 2 880 640 22 40 32 20 32 32 32 8 1 4 3 2 880 640 22 56 48 20 48 128 256 8 1 19 1 1 1232 960 22 56 48 20 48 64 64 8 1 10 1 1 1232 960 12 40 32 10 32 256 512 8 1 32 1 1 480 320 12 40 32 10 32 128 128 8 1 16 1 1 480 320 4 48 48 4 48 256 256 8 1 8 1 5 192 192 10 24 24 10 24 512 256 32 1 16 1 1 240 240 5 12 12 5 12 512 256 32 1 16 1 1 60 60 3 6 6 3 6 512 256 32 1 16 1 1 18 18 2 3 3 2 3 512 256 32 1 16 1 1 6 6 4 48 48 4 48 256 16 8 1 8 1 5 192 192 4 48 48 4 48 256 88 8 1 8 1 5 192 192 10 24 24 10 24 256 24 24 1 8 1 1 240 240 10 24 24 10 24 256 128 32 1 8 1 1 240 240 5 12 12 5 12 256 24 24 1 8 1 1 60 60 5 12 12 5 12 256 128 32 1 8 1 1 60 60 3 6 6 3 6 256 24 24 1 8 1 1 18 18 3 6 6 3 6 256 128 32 1 8 1 1 18 18 2 3 3 2 3 256 16 16 1 8 1 1 6 6 2 3 3 2 3 256 96 32 1 8 1 1 6 6 254 , TIDL_BatchNormLayer , PASSED #MMACs = 0.74, 0.74, Sparsity : 0.00 4787 , TIDL_ConvolutionLayer, PASSED #MMACs = 147.46, 92.65, Sparsity : 37.17 4230 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 53.33, Sparsity : 62.33 7280 , TIDL_ConvolutionLayer, PASSED #MMACs = 283.12, 83.44, Sparsity : 70.53 10223 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 66.11, Sparsity : 53.30 8988 , TIDL_ConvolutionLayer, PASSED #MMACs = 283.12, 91.59, Sparsity : 67.65 10923 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 57.32, Sparsity : 59.51 20852 , TIDL_ConvolutionLayer, PASSED #MMACs = 283.12, 96.27, Sparsity : 66.00 18101 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 52.28, Sparsity : 63.07 PASSED #MMACs = 0.06, 0.06, Sparsity : 0.00 27171 , TIDL_ConvolutionLayer, PASSED #MMACs = 283.12, 76.31, Sparsity : 73.04 5405 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 31.40, Sparsity : 77.82 PASSED #MMACs = 0.03, 0.03, Sparsity : 0.00 PASSED #MMACs = 0.01, 0.01, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 15924 , TIDL_ConvolutionLayer, PASSED #MMACs = 62.91, 62.91, Sparsity : 0.00 10177 , TIDL_ConvolutionLayer, PASSED #MMACs = 31.46, 31.46, Sparsity : 0.00 14028 , TIDL_ConvolutionLayer, PASSED #MMACs = 7.86, 7.86, Sparsity : 0.00 17569 , TIDL_ConvolutionLayer, PASSED #MMACs = 2.36, 2.36, Sparsity : 0.00 26121 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.79, 0.79, Sparsity : 0.00 4366 , TIDL_ConvolutionLayer, PASSED #MMACs = 3.93, 3.93, Sparsity : 0.00 PASSED #MMACs = 0.02, 0.02, Sparsity : 0.00 3862 , TIDL_ConvolutionLayer, PASSED #MMACs = 21.63, 21.63, Sparsity : 0.00 PASSED #MMACs = 0.08, 0.08, Sparsity : 0.00 5460 , TIDL_ConvolutionLayer, PASSED #MMACs = 1.47, 1.47, Sparsity : 0.00 PASSED #MMACs = 0.01, 0.01, Sparsity : 0.00 2597 , TIDL_ConvolutionLayer, PASSED #MMACs = 7.86, 7.86, Sparsity : 0.00 PASSED #MMACs = 0.03, 0.03, Sparsity : 0.00 6983 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.37, 0.37, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 2508 , TIDL_ConvolutionLayer, PASSED #MMACs = 1.97, 1.97, Sparsity : 0.00 PASSED #MMACs = 0.01, 0.01, Sparsity : 0.00 9470 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.11, 0.11, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 3264 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.59, 0.59, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 8417 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.02, 0.02, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 3940 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.15, 0.15, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 4383 , TIDL_ConcatLayer, PASSED #MMACs = 0.00, 0.00, Sparsity : -1.#J 2518 , TIDL_ConcatLayer, PASSED #MMACs = 0.00, 0.00, Sparsity : -1.#J 0.00, 0.00, Sparsity : 0.00 _03_03_00_00\ti_components\algorithms\REL.TIDLSRC.01.01.02.00\modules\ti_dl\utils\tidlModel = j('#fragment-1a84591e31034fac832d29ed8584666c1664508081_code-editor-fs'); = j('#fragment-1a84591e31034fac832d29ed8584666c1664508081_code-editor'); function(){ evolutionCodeEditor('fullscreen')) { evolutionCodeEditor('fullscreen', false); evolutionCodeEditor('fullscreen', true);

What else should I check?

Thank you for your help.

BR,

Khethan

0 Praveen Eppa1 over 7 years ago in reply to khethan

TI__Genius 17580 points

Hi Khethan,

Can you share your output ?

Thanks,
Praveen

0 khethan over 7 years ago in reply to Praveen Eppa1

Intellectual 470 points

Hi Praveen,

Here I attach the output for the input.

tempDir.zip

BR,

Khethan

0 khethan over 7 years ago in reply to khethan

Intellectual 470 points

Hi Praveen,

I have tested on TDA2P EVM, but the output data is initialized as below attachment files.

evm_output.zip

This phenomenon seems to be the same in PC simulation mode.

As mentioned earlier, if I use eve_test_dl_algo.out.exe, which is included in the package by default, the output is very good.

The TIDL build was referenced in the TIDeepLearningLibrary_UserGuide.pdf document.

BR,

Khethan

0 khethan over 7 years ago in reply to khethan

Intellectual 470 points

Hi Praveen,

I tried to compile eve and dsp again with TIDL_SRC(1.1.2), and confirmed that the generated file (dsp_test_dl_algo.out, eve_test_dl_algo.out) works fine on EVM through CCS as output file(stats_tool_out.bin).

However, when I execute the same model file with "tidl od usecase" in EVM(sd_boot), the video is played but the bbox is not displayed.

please refer my data.

eve_data_log.zip

I would appreciate your advice on what went wrong.

BR,

Khethan

0 Praveen Eppa1 over 7 years ago in reply to khethan

TI__Genius 17580 points

Hi Khethan,

Were you able get TIDL Usecase working as is from release package?

Thanks,
Praveen

0 khethan over 7 years ago in reply to Praveen Eppa1

Intellectual 470 points

Hi Praveen,

The package(3.3.0.0) which i use current was not supported TIDL usecase for TDA2P, but I confirmed that it works well after modifying the chain etc with reference to TDA2X.

Regarding above issue, It was related to layersGroupId valule in tidl_import_JDetNex.txt.

When I values just change as below regardless of version, the problem was solved.

before:

layersGroupId = 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 0

conv2dKernelType = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

after:

layersGroupId = 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 2 2 2 0
conv2dKernelType = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

But there was another problem.

1. When tidl usecase is executed, bbox is not displayed from the beginning and box is displayed from about 20sec of video.

2. When video(inData and inHeader) are replaced with mine, frame is played at 1 ~ 2fps slowly, resulting in frame reversal. the Problem is not reproduced when using ti demo clip.

I made this clip using the program(ffmpeg&ffprobe&sizeBin) and It seem like to have no problem.

BR,

Khethan

0 khethan over 7 years ago in reply to khethan

Intellectual 470 points

Hi Praveen

I have reduced the time to 6sec by adjusting parameters( quantHistoryParam1 / 2 and quantMargin) as below.

20, 5, 0 -> 40, 40, 40

Is there any problem using this value? How can we reduce the lead time further?

I have read for the doc(tidl user guide) for parameter but I do not understand it enough.

could you please explain about the parameter in detail(unit.etc)

The attached log file does not seem to be a problem.

revers_sym_log.txt

Referencing TIDeepLearningLibrary_UserGuide.pdf 3.7 (Input and Output Data Formats), clip format is below. right?

width: 760(video)+4(MAX_PAD), height; 312(video)+4(MAX_PAD)

If it is not related to it, What should I do ?

BR,

Khethan

0 khethan over 7 years ago in reply to khethan

Intellectual 470 points

I am sorry,

The clip size should be modified as follows due to my mistake
width: 760(video)+2*4(MAX_PAD), height; 312(video)+2*4(MAX_PAD)

BR,
Khethan

0 Praveen Eppa1 over 7 years ago in reply to khethan

TI__Genius 17580 points

Hi Khethan,

#1

The lead time can not be reduced by changing these quant parameters, these parameters are used to improve the accuracy, I explained these parameters in detail below,

In TIDL, we use the current computation of min and max to update the quantization parameters for next frame. We don’t directly use it in the next frame, we gradually update.
This quantHistoryParam1 / 2 and quantMargin values are used to control, how fast we need update the quantization parameter.
If quantHistoryParam1 / 2 are higher than the update will happen faster.
QuantMargin controls the margin that would want for max to grow.

#2

Yes, this padding is correct

Thanks,
Praveen

0 khethan over 7 years ago in reply to Praveen Eppa1

Intellectual 470 points

Hi Praveen,

My issue was solved
Thank you for your support.

BR,
Khethan

Processors

Processors forum

TDA2: ssd model import problem