1. 03 Mar, 2020 1 commit
    • Liangzhe's avatar
      Merged commit includes the following changes: (#8235) · 4a0b3e4b
      Liangzhe authored
      
      
      298416930  by lzyuan:
      
          Explicitly mark base models' state inputs as 'raw_inputs/init_lstm_h_1' and 'raw_inputs_init_lstm_h_2' when pre_bottleneck=True.
      
      --
      298380851  by skligys:
      
          Fix LSTD LSTM cells to use fixed_quantize_op().
      
      --
      297662737  by Menglong Zhu:
      
          Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration
      
      --
      289667197  by lzyuan:
      
          Internal update.
      
      --
      288607438  by lzyuan:
      
          Enforce feature_extractor construction using arg keys.
      
      --
      
      PiperOrigin-RevId: 298416930
      Co-authored-by: default avatarMenglong Zhu <menglong@google.com>
      4a0b3e4b
  2. 12 Aug, 2019 1 commit
    • Richard Brooks's avatar
      Lstm object detection improvements (#7379) · 49075e50
      Richard Brooks authored
      * Replace google3.pyglib modules with tf and absl
      
      This now matches train.py and provides more publicly available libraries.
      
      * Add example pipeline config for SSD Interleaved V2 Model.
      
      Compiled from model_builder_test.py and lstm_ssd_mobilenet_v1_imagenet.config,
      Removed data augmentation and tranfer learning (i.e. training from checkpoint) due to errors I was seeing when trying to run with it.
      
      * Remove unused tfrecord creation.
      
      This was also incorrectly specified, as the keys differed from the TFSequenceExample parser.
      
      * correct key specified in docstring
      
      * add tflite frozen graph exporter (cli and lib).
      
      * add tflite model exporter
      
      * add script to test the tflite model
      
      * add mode export documentation
      
      * correct docstring
      
      * rename export files to be unique across detection research work
      
      * correct number of channels for grayscale
      
      * add and correct copyright
      49075e50
  3. 03 Dec, 2018 1 commit
  4. 26 Nov, 2018 2 commits
    • Menglong Zhu's avatar
      Fix dependency issues (#5815) · 2d9d7943
      Menglong Zhu authored
      * Add a second expected_loss_weights function using an alternative expectation calculation compared to previous. Integrate this op into ssd_meta_arch and losses builder. Affects files that use losses_builder.build to handle the returning of an additional element.
      
      PiperOrigin-RevId: 219190083
      
      * Fix dependency issues.
      
      PiperOrigin-RevId: 222888231
      
      * Update CODEOWNERS 
      
      Add @masonliuw and @yinxiaoli to /research/lstm_object_detection/
      2d9d7943
    • Menglong Zhu's avatar
      Add a second expected_loss_weights function using an alternative expectation... · 8ab3e7a0
      Menglong Zhu authored
      Add a second expected_loss_weights function using an alternative expectation calculation compared to previous. Integrate this op into ssd_meta_arch and losses builder. Affects files that use losses_builder.build to handle the returning of an additional element.
      
      PiperOrigin-RevId: 219190083
      8ab3e7a0
  5. 24 Oct, 2018 1 commit