Commit Graph

  • 225a720f18 updated README to include testing master Malar Kannan 2017-12-29 16:21:38 +0530
  • b267b89a44 Merge branch 'master' of /home/ilml/Public/Repos/speech_scoring Malar Kannan 2017-12-29 13:15:51 +0530
  • eb10b577ae Added README.md describing the workflow Malar Kannan 2017-12-29 13:14:37 +0530
  • ee2eb63f66 Merge branch 'master' of ssh://invnuc/~/Public/Repos/speech_scoring Malar Kannan 2017-12-28 20:02:44 +0530
  • 2ae269d939 generating test for phone seg model Malar Kannan 2017-12-28 20:01:44 +0530
  • 40d7933870 saving model on better 'acc' Malar Kannan 2017-12-28 20:00:19 +0530
  • 4dd4bb5963 implemented phoneme segmented training on samples Malar Kannan 2017-12-28 18:53:54 +0530
  • 0600482fe5 generating segmentation for words Malar Kannan 2017-12-28 13:37:27 +0530
  • 507da49cfa added voicerss tts support for test data generation Malar Kannan 2017-12-26 14:32:56 +0530
  • f44665e9b2 1. fixed softmax output and overfit the model for small sample 2. updated to run on complete data Malar Kannan 2017-12-12 11:38:27 +0530
  • cc4fbe45b9 trying to overfit 2 samples with model -> doesn't seem to converge Malar Kannan 2017-12-11 15:03:14 +0530
  • 8d550c58cc fixed batch normalization layer before activation Malar Kannan 2017-12-11 14:32:39 +0530
  • 240ecb3f27 removed bn output layer Malar Kannan 2017-12-11 14:12:23 +0530
  • 05242d5991 added batch normalization Malar Kannan 2017-12-11 14:09:04 +0530
  • fea9184aec using the full data and fixed typo in model layer name Malar Kannan 2017-12-11 13:47:30 +0530
  • a6543491f8 fixed empty phoneme boundary case Malar Kannan 2017-12-11 13:05:46 +0530
  • d387922f7d added dense-relu/softmax layers to segment output Malar Kannan 2017-12-11 12:30:08 +0530
  • 52bbb69c65 resuming segment training Malar Kannan 2017-12-10 21:58:55 +0530
  • 03edd935ea fixed input_dim Malar Kannan 2017-12-07 17:15:44 +0530
  • a7f1451a7f fixed exception in data generation Malar Kannan 2017-12-07 16:49:34 +0530
  • 91fde710f3 completed the segmentation model Malar Kannan 2017-12-07 15:17:59 +0530
  • c8a07b3d7b Merge branch 'master' of ssh://invnuc/~/Public/Repos/speech_scoring Malar Kannan 2017-12-07 12:00:59 +0530
  • 8785522196 Merge branch 'master' of /home/ilml/Public/Repos/speech_scoring Malar Kannan 2017-12-07 12:00:44 +0530
  • 435c4a4aa6 added a resume parameter for training Malar Kannan 2017-12-07 12:00:42 +0530
  • c1801b5aa3 implented segment tfrecords batch data-generator Malar Kannan 2017-12-07 11:48:19 +0530
  • c0369d7a66 Merge branch 'master' of ssh://gpuaws/~/repos/speech_scoring Malar Kannan 2017-12-06 17:33:27 +0530
  • 8e14db2437 Merge branch 'master' of ssh://invmac/~/Public/repos/speech-scoring Malar Kannan 2017-12-06 17:32:46 +0530
  • bcf1041bde created segment sample tfrecord writer Malar Kannan 2017-12-06 17:32:26 +0530
  • b50edb980d implemented segment-generation for random words for testing Malar Kannan 2017-12-06 14:41:25 +0530
  • 3f76207f0d using pitch contour instead of spectrogram Malar Kannan 2017-12-04 19:15:17 +0530
  • 6ef4e86f41 implemented segmentation visualization Malar Kannan 2017-11-30 14:49:55 +0530
  • 0b1152b5c3 implemented the model, todo implement ctc and training queueing logic Malar Kannan 2017-11-28 19:10:19 +0530
  • 1928fce4e8 Merge branch 'master' of ssh://invnuc/~/Public/Repos/speech_scoring Malar Kannan 2017-11-28 17:05:35 +0530
  • ec7303223c merged Malar Kannan 2017-11-28 17:05:20 +0530
  • f12da988d3 segmentation model wip Malar Kannan 2017-11-28 15:46:39 +0530
  • 705cf3d172 finding exact duration of sound sample Malar Kannan 2017-11-28 12:52:00 +0530
  • 8f79316893 Merge branch 'master' of /Users/malarkannan/Public/repos/speech-scoring Malar Kannan 2017-11-28 12:32:50 +0530
  • 0345cc46ae implemented tts sementation generation code Malar Kannan 2017-11-28 12:16:57 +0530
  • 20b2d7a958 updated model data Malar Kannan 2017-11-27 14:08:01 +0530
  • 43d5b75db9 removing spec_n counter Malar Kannan 2017-11-24 11:06:42 +0000
  • ec08cc7d62 Merge branch 'master' of ssh://gpuaws/~/repos/speech_scoring Malar Kannan 2017-11-24 14:32:43 +0530
  • 2268ad8bb0 implemented pitch plotting Malar Kannan 2017-11-24 14:32:13 +0530
  • ec317b6628 Merge branch 'master' of /home/ilml/Public/Repos/speech_scoring Malar Kannan 2017-11-24 14:26:40 +0530
  • 235300691e find spec_n from tfrecords Malar Kannan 2017-11-24 14:26:36 +0530
  • ae46578aec Merge branch 'master' of ssh://invmac/~/Public/repos/speech-scoring Malar Kannan 2017-11-23 17:50:47 +0530
  • 3d7542271d implemented tts segmentation data generation Malar Kannan 2017-11-23 17:50:11 +0530
  • 54f38ca775 removed a layer using lstm Malar Kannan 2017-11-22 15:46:42 +0530
  • 6355db4af7 adding missing model-dir for training constants copying Malar Kannan 2017-11-22 15:04:02 +0530
  • 1f60183ab8 Merge branch 'master' of ssh://invnuc/~/Public/Repos/speech_scoring Malar Kannan 2017-11-22 14:45:35 +0530
  • e7fc607578 trying mfcc instead of spectrogram Malar Kannan 2017-11-22 14:45:08 +0530
  • d2a075422c copying constantws to models Malar Kannan 2017-11-20 15:15:27 +0530
  • a5d4ede35d finding number of record by streaming-onepass Malar Kannan 2017-11-20 12:07:13 +0530
  • 3ae8dc50a2 implemented pair data inspection Malar Kannan 2017-11-17 17:29:48 +0530
  • c81a7b4468 decreasing first layer node count to avoid gpu memory overflow Ubuntu 2017-11-17 10:31:36 +0000
  • c682962c8f using a Bi-LSTM layer as the first layer Malar Kannan 2017-11-17 14:17:12 +0530
  • 6ff052be9b fixed randomize pair picking Malar Kannan 2017-11-17 11:57:38 +0530
  • 7fc89c0853 1. fixed pairing and data duplicates 2. clean-up Malar Kannan 2017-11-16 22:56:24 +0530
  • 3d297f176f perfect score on new test words - TODO evaluate on real voice Malar Kannan 2017-11-16 14:19:25 +0530
  • 7d94ddc2ae all phrases Malar Kannan 2017-11-15 18:30:43 +0530
  • 77c7adbdb5 Merge branch 'master' of ssh://invmac/~/Public/repos/speech-scoring Malar Kannan 2017-11-15 18:29:23 +0530
  • a67ce148d6 fixed dupliate words Malar Kannan 2017-11-15 18:27:49 +0530
  • c75ff4d109 failure visualization wip Malar Kannan 2017-11-15 15:17:37 +0530
  • a9b244a50c the pair generation order is randomized Malar Kannan 2017-11-15 14:43:39 +0530
  • 1b0ba26a6e Merge branch 'master' of ssh://invnuc/~/Public/Repos/speech_scoring Malar Kannan 2017-11-15 14:17:15 +0530
  • e9f54c7f6f 1. tuned batchsize 2. fixed last batch carry-over Malar Kannan 2017-11-15 14:14:17 +0530
  • 7684ab3a74 ported to tqdm Malar Kannan 2017-11-14 22:56:13 +0530
  • 036667d1c7 Merge branch 'master' of https://code.whiteblossom.net/malar/speech-scoring Malar Kannan 2017-11-14 21:59:15 +0530
  • 10b024866e implemented evaluation of test data with model by overfitting on smaller dataset Malar Kannan 2017-11-14 17:54:44 +0530
  • e4b8b4e0a7 visualizing and playing sound files where prediction fails Malar Kannan 2017-11-13 19:22:30 +0530
  • 988f66c2c2 avoiding same voice similar variants Malar Kannan 2017-11-13 17:33:37 +0530
  • d978272bdb saving model and tensorboard Malar Kannan 2017-11-10 18:06:45 +0530
  • bb72c4045e trying to overfit the model to identify false-negative types Malar Kannan 2017-11-10 17:52:21 +0530
  • 1190312def removed tfrecord tensor code and remnants Malar Kannan 2017-11-10 14:15:12 +0530
  • e9b18921ee implemented train/test split at word-level and generator returns one-shot validation data Malar Kannan 2017-11-10 14:07:31 +0530
  • ab452494b3 implemented streaming tfreccords Malar Kannan 2017-11-09 20:31:29 +0530
  • 0a4d4fadeb implemented random sampling of data for oneshot loading Malar Kannan 2017-11-09 15:00:17 +0530
  • b3a6aa2f6a clean-up Malar Kannan 2017-11-08 11:08:19 +0530
  • 7cbfebbf1a 1. fixed missing wrong pairs 2.using different progress bakend Malar Kannan 2017-11-07 17:27:01 +0530
  • b8a9f87031 implemented padding and pipeline is complete Malar Kannan 2017-11-07 15:18:04 +0530
  • 41b3f1a9fe dropping invalid csv entries Malar Kannan 2017-11-07 12:43:17 +0530
  • 55e2de2f04 using csv writer instead as comma in phrases are mis-aligning columns Malar Kannan 2017-11-07 11:56:09 +0530
  • 33c6bcc3c1 implemeted test data sample generation Malar Kannan 2017-11-07 10:23:31 +0530
  • 15f29895d4 implemented tfrecord reader and model refactor wip Malar Kannan 2017-11-07 00:10:23 +0530
  • 5b682c78b8 Merge branch 'master' of ssh://invmac/~/Public/repos/speech-scoring Malar Kannan 2017-11-06 15:50:06 +0530
  • 046343680e implemented siamese pair tfrecord writer Malar Kannan 2017-11-06 15:48:38 +0530
  • c187fbe1ca implemented tfrecord writer for spectrograms Malar Kannan 2017-11-06 14:12:09 +0530
  • fabd882664 tfrecords wip Malar Kannan 2017-11-06 12:36:20 +0530
  • 5ff437b095 computing spectrogram for existing files Malar Kannan 2017-11-06 12:15:12 +0530
  • 4194e05b4c removing - from phrases before synthesizing audio Malar Kannan 2017-11-03 15:30:13 +0530
  • 22d353f101 skipping missing files Malar Kannan 2017-11-03 15:20:31 +0530
  • 1f19463b65 computing phoneme/word variant for each word in a phrase Malar Kannan 2017-11-03 14:48:55 +0530
  • b4ceeb4eed generating spectrogram parallelly Malar Kannan 2017-11-03 14:19:19 +0530
  • 6ab84b4dc2 Merge branch 'master' of ssh://invmac/~/Public/repos/speech-scoring Malar Kannan 2017-11-02 13:16:04 +0530
  • d4454b6434 looping record test code Malar Kannan 2017-11-02 13:14:59 +0530
  • 45977a819d generating randome samples Malar Kannan 2017-11-02 13:14:08 +0530
  • 4188585488 updated test code Malar Kannan 2017-10-31 17:41:02 +0530
  • 6fbf06814c updated model to use dense classifier Malar Kannan 2017-10-31 13:31:31 +0530
  • 2d9b12af95 fixed out of range exception Malar Kannan 2017-10-31 10:29:24 +0530
  • 80c0ce403e generating all words for a every voice first Malar Kannan 2017-10-27 19:04:09 +0530
  • cbf15ff662 type in fn name Malar Kannan 2017-10-27 18:57:26 +0530